Datasets:

---

ftopal

/

huggingface-datasets

like 0

[ "license:cc-by-nc-4.0", "region:us" ]

[]

[]

[ "# Vietnamese Text-To-Speech dataset (VietTTS-v1.1)\n visit URL for a vietnamese TTS library (included pretrained models). \n\nThe text is from a collection of novels and short stories from the author \"Vu Trong Phung.\" The text is in public domain.\nThe audio is generated by Google Text-to-Speech offline engine on Android. The audio is NOT for commercial use.\n\nDataset size: '5.4G'.\nTotal audio duration: '35.9 hours'.", "### Text-audio samples\n\n - Sample 1: \n + Audio: file1\n + Text: '\"Ai\" đây tức là một kẻ ăn mày vậy. Anh ta chưa kịp quay đi thì đã thấy mấy con chó vàng chạy xồng xộc ra cứ nhảy xổ vào chân anh.'\n\n - Sample 2:\n + Audio: file2\n + Text: 'Ừ, thế mày đã nuôi được bố mẹ mày bữa nào chưa, hay xưa nay vẫn báo hại cơm cha áo mẹ mãi? Mấy hôm thấy ông đơ mặt không thèm nói, mày lại làm già à?'", "### Download\nGet the dataset from here: link.\n Or, run the following commands: \n\n\n\n'dataset' directory structure:", "### Statistics\n\n - Number of clips: 22884 clips.\n - Shortest audio clip: 0.46 seconds.\n - Median clip duration: 5.46 seconds.\n - Mean clip duration: 5.65 seconds.\n - Longest audio clip: 15.4 seconds.", "### Vũ Trọng Phụng's collections\n\n- Bệnh Lao Chữa Bằng Mồm Hay Là ... Thầy Lang Bất Hủ, 1934?\n- Cạm Bẫy Người, 1933.\n- Cơm Thầy Cơm Cô, 1936.\n- Đời Là Một Cuộc Chiến Đấu,1939.\n- Dứt Tình, 1934.\n- Giông Tố, 1936.\n- Gương Tống Tiền, N/A.\n- Hồ Sê Líu, Hồ Líu Sê Sàng, 1936.\n- Kỹ Nghệ Lấy Tây, 1934.\n- Làm Đĩ, 1936.\n- Lấy Nhau Vì Tình, 1937.\n- Lấy Vợ Xấu, 1937.\n- Lòng Tự Ái, 1937.\n- Máu Mê, 1937.\n- Một Cái Chết, 1931.\n- Một Con Chó Hay Chim Chuột, 1937.\n- Một Đồng Bạc, 1939.\n- Người Có Quyền, 1937.\n- Sao Mày Không Vỡ Nắp Ơi!, 1934.\n- Số Đỏ, 1936.\n- Sư Cụ Triết Lý, 1935.\n- Trúng Số Độc Đắc, 1938.\n- Tự Do, 1937.\n- Từ Lý Thuyết Đến Thực Hành, N/A.\n- Vỡ Đê, 1936." ]

[ "TAGS\n#license-cc-by-nc-4.0 #region-us \n", "# Vietnamese Text-To-Speech dataset (VietTTS-v1.1)\n visit URL for a vietnamese TTS library (included pretrained models). \n\nThe text is from a collection of novels and short stories from the author \"Vu Trong Phung.\" The text is in public domain.\nThe audio is generated by Google Text-to-Speech offline engine on Android. The audio is NOT for commercial use.\n\nDataset size: '5.4G'.\nTotal audio duration: '35.9 hours'.", "### Text-audio samples\n\n - Sample 1: \n + Audio: file1\n + Text: '\"Ai\" đây tức là một kẻ ăn mày vậy. Anh ta chưa kịp quay đi thì đã thấy mấy con chó vàng chạy xồng xộc ra cứ nhảy xổ vào chân anh.'\n\n - Sample 2:\n + Audio: file2\n + Text: 'Ừ, thế mày đã nuôi được bố mẹ mày bữa nào chưa, hay xưa nay vẫn báo hại cơm cha áo mẹ mãi? Mấy hôm thấy ông đơ mặt không thèm nói, mày lại làm già à?'", "### Download\nGet the dataset from here: link.\n Or, run the following commands: \n\n\n\n'dataset' directory structure:", "### Statistics\n\n - Number of clips: 22884 clips.\n - Shortest audio clip: 0.46 seconds.\n - Median clip duration: 5.46 seconds.\n - Mean clip duration: 5.65 seconds.\n - Longest audio clip: 15.4 seconds.", "### Vũ Trọng Phụng's collections\n\n- Bệnh Lao Chữa Bằng Mồm Hay Là ... Thầy Lang Bất Hủ, 1934?\n- Cạm Bẫy Người, 1933.\n- Cơm Thầy Cơm Cô, 1936.\n- Đời Là Một Cuộc Chiến Đấu,1939.\n- Dứt Tình, 1934.\n- Giông Tố, 1936.\n- Gương Tống Tiền, N/A.\n- Hồ Sê Líu, Hồ Líu Sê Sàng, 1936.\n- Kỹ Nghệ Lấy Tây, 1934.\n- Làm Đĩ, 1936.\n- Lấy Nhau Vì Tình, 1937.\n- Lấy Vợ Xấu, 1937.\n- Lòng Tự Ái, 1937.\n- Máu Mê, 1937.\n- Một Cái Chết, 1931.\n- Một Con Chó Hay Chim Chuột, 1937.\n- Một Đồng Bạc, 1939.\n- Người Có Quyền, 1937.\n- Sao Mày Không Vỡ Nắp Ơi!, 1934.\n- Số Đỏ, 1936.\n- Sư Cụ Triết Lý, 1935.\n- Trúng Số Độc Đắc, 1938.\n- Tự Do, 1937.\n- Từ Lý Thuyết Đến Thực Hành, N/A.\n- Vỡ Đê, 1936." ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for \"github-issues\"\n\nMore Information needed" ]

[ "TAGS\n#region-us \n", "# Dataset Card for \"github-issues\"\n\nMore Information needed" ]

[ "task_categories:question-answering", "task_categories:multiple-choice", "task_ids:multiple-choice-qa", "task_ids:open-domain-qa", "annotations_creators:no-annotation", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:apache-2.0", "region:us" ]

[]

[ "en" ]

[ "### Dataset Summary\n\n\nMedMCQA is a large-scale, Multiple-Choice Question Answering (MCQA) dataset designed to address real-world medical entrance exam questions.\n\n\nMedMCQA has more than 194k high-quality AIIMS & NEET PG entrance exam MCQs covering 2.4k healthcare topics and 21 medical subjects are collected with an average token length of 12.77 and high topical diversity.\n\n\nEach sample contains a question, correct answer(s), and other options which require a deeper language understanding as it tests the 10+ reasoning abilities of a model across a wide range of medical subjects & topics. A detailed explanation of the solution, along with the above information, is provided in this study.\n\n\nMedMCQA provides an open-source dataset for the Natural Language Processing community.\nIt is expected that this dataset would facilitate future research toward achieving better QA systems.\nThe dataset contains questions about the following topics:\n\n\n* Anesthesia\n* Anatomy\n* Biochemistry\n* Dental\n* ENT\n* Forensic Medicine (FM)\n* Obstetrics and Gynecology (O&G)\n* Medicine\n* Microbiology\n* Ophthalmology\n* Orthopedics\n* Pathology\n* Pediatrics\n* Pharmacology\n* Physiology\n* Psychiatry\n* Radiology\n* Skin\n* Preventive & Social Medicine (PSM)\n* Surgery", "### Supported Tasks and Leaderboards\n\n\nmultiple-choice-QA, open-domain-QA: The dataset can be used to train a model for multi-choice questions answering, open domain questions answering. Questions in these exams are challenging and generally require deeper domain and language understanding as it tests the 10+ reasoning abilities across a wide range of medical subjects & topics.", "### Languages\n\n\nThe questions and answers are available in English.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'id' : a string question identifier for each example\n* 'question' : question text (a string)\n* 'opa' : Option A\n* 'opb' : Option B\n* 'opc' : Option C\n* 'opd' : Option D\n* 'cop' : Correct option, i.e., 1,2,3,4\n* 'choice\\_type' ({\"single\", \"multi\"}): Question choice type.\n\t+ \"single\": Single-choice question, where each choice contains a single option.\n\t+ \"multi\": Multi-choice question, where each choice contains a combination of multiple suboptions.\n* 'exp' : Expert's explanation of the answer\n* 'subject\\_name' : Medical Subject name of the particular question\n* 'topic\\_name' : Medical topic name from the particular subject", "### Data Splits\n\n\nThe goal of MedMCQA is to emulate the rigor of real word medical exams. To enable that, a predefined split of the dataset is provided. The split is by exams instead of the given questions. This also ensures the reusability and generalization ability of the models.\n\n\nThe training set of MedMCQA consists of all the collected mock & online test series, whereas the test set consists of all AIIMS PG exam MCQs (years 1991-present). The development set consists of NEET PG exam MCQs (years 2001-present) to approximate real exam evaluation.\n\n\nSimilar questions from train , test and dev set were removed based on similarity. The final split sizes are as follow:\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nBefore this attempt, very few works have been done to construct biomedical MCQA datasets (Vilares and Gomez-Rodr, 2019), and they are (1) mostly small, containing up to few thousand questions, and (2) cover a limited number of Medical topics and Subjects. This paper addresses the aforementioned limitations by introducing MedMCQA, a new large-scale, Multiple-Choice Question Answering\n(MCQA) dataset designed to address real-world medical entrance exam questions.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nHistorical Exam questions from official websites - AIIMS & NEET PG (1991- present)\nThe raw data is collected from open websites and books", "#### Who are the source language producers?\n\n\nThe dataset was created by Ankit Pal, Logesh Kumar Umapathi and Malaikannan Sankarasubbu", "### Annotations", "#### Annotation process\n\n\nThe dataset does not contain any additional annotations.", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nIf you find this useful in your research, please consider citing the dataset paper", "### Contributions\n\n\nThanks to @monk1337 for adding this dataset." ]

[ "TAGS\n#task_categories-question-answering #task_categories-multiple-choice #task_ids-multiple-choice-qa #task_ids-open-domain-qa #annotations_creators-no-annotation #language_creators-expert-generated #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-apache-2.0 #region-us \n", "### Dataset Summary\n\n\nMedMCQA is a large-scale, Multiple-Choice Question Answering (MCQA) dataset designed to address real-world medical entrance exam questions.\n\n\nMedMCQA has more than 194k high-quality AIIMS & NEET PG entrance exam MCQs covering 2.4k healthcare topics and 21 medical subjects are collected with an average token length of 12.77 and high topical diversity.\n\n\nEach sample contains a question, correct answer(s), and other options which require a deeper language understanding as it tests the 10+ reasoning abilities of a model across a wide range of medical subjects & topics. A detailed explanation of the solution, along with the above information, is provided in this study.\n\n\nMedMCQA provides an open-source dataset for the Natural Language Processing community.\nIt is expected that this dataset would facilitate future research toward achieving better QA systems.\nThe dataset contains questions about the following topics:\n\n\n* Anesthesia\n* Anatomy\n* Biochemistry\n* Dental\n* ENT\n* Forensic Medicine (FM)\n* Obstetrics and Gynecology (O&G)\n* Medicine\n* Microbiology\n* Ophthalmology\n* Orthopedics\n* Pathology\n* Pediatrics\n* Pharmacology\n* Physiology\n* Psychiatry\n* Radiology\n* Skin\n* Preventive & Social Medicine (PSM)\n* Surgery", "### Supported Tasks and Leaderboards\n\n\nmultiple-choice-QA, open-domain-QA: The dataset can be used to train a model for multi-choice questions answering, open domain questions answering. Questions in these exams are challenging and generally require deeper domain and language understanding as it tests the 10+ reasoning abilities across a wide range of medical subjects & topics.", "### Languages\n\n\nThe questions and answers are available in English.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'id' : a string question identifier for each example\n* 'question' : question text (a string)\n* 'opa' : Option A\n* 'opb' : Option B\n* 'opc' : Option C\n* 'opd' : Option D\n* 'cop' : Correct option, i.e., 1,2,3,4\n* 'choice\\_type' ({\"single\", \"multi\"}): Question choice type.\n\t+ \"single\": Single-choice question, where each choice contains a single option.\n\t+ \"multi\": Multi-choice question, where each choice contains a combination of multiple suboptions.\n* 'exp' : Expert's explanation of the answer\n* 'subject\\_name' : Medical Subject name of the particular question\n* 'topic\\_name' : Medical topic name from the particular subject", "### Data Splits\n\n\nThe goal of MedMCQA is to emulate the rigor of real word medical exams. To enable that, a predefined split of the dataset is provided. The split is by exams instead of the given questions. This also ensures the reusability and generalization ability of the models.\n\n\nThe training set of MedMCQA consists of all the collected mock & online test series, whereas the test set consists of all AIIMS PG exam MCQs (years 1991-present). The development set consists of NEET PG exam MCQs (years 2001-present) to approximate real exam evaluation.\n\n\nSimilar questions from train , test and dev set were removed based on similarity. The final split sizes are as follow:\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nBefore this attempt, very few works have been done to construct biomedical MCQA datasets (Vilares and Gomez-Rodr, 2019), and they are (1) mostly small, containing up to few thousand questions, and (2) cover a limited number of Medical topics and Subjects. This paper addresses the aforementioned limitations by introducing MedMCQA, a new large-scale, Multiple-Choice Question Answering\n(MCQA) dataset designed to address real-world medical entrance exam questions.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nHistorical Exam questions from official websites - AIIMS & NEET PG (1991- present)\nThe raw data is collected from open websites and books", "#### Who are the source language producers?\n\n\nThe dataset was created by Ankit Pal, Logesh Kumar Umapathi and Malaikannan Sankarasubbu", "### Annotations", "#### Annotation process\n\n\nThe dataset does not contain any additional annotations.", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nIf you find this useful in your research, please consider citing the dataset paper", "### Contributions\n\n\nThanks to @monk1337 for adding this dataset." ]

[ "task_categories:audio-classification", "annotations_creators:unknown", "language_creators:unknown", "size_categories:10K<n<100K", "source_datasets:unknown", "license:cc-by-4.0", "arxiv:2010.00475", "region:us" ]

[ "2010.00475" ]

[]

[ "# Freesound Dataset 50k (FSD50K)", "## Important\n\nThis data set is a copy from the original one located at Zenodo.", "## Dataset Description\n- Homepage: FSD50K\n- Repository: GitHub\n- Paper: FSD50K: An Open Dataset of Human-Labeled Sound Events\n- Leaderboard: Paperswithcode Leaderboard\n\nIf you use the FSD50K dataset, or part of it, please cite our paper:\n\n>Eduardo Fonseca, Xavier Favory, Jordi Pons, Frederic Font, Xavier Serra. \"FSD50K: an Open Dataset of Human-Labeled Sound Events\", arXiv 2020.", "### Data curators\n\nEduardo Fonseca, Xavier Favory, Jordi Pons, Mercedes Collado, Ceren Can, Rachit Gupta, Javier Arredondo, Gary Avendano and Sara Fernandez", "### Contact\n\nYou are welcome to contact Eduardo Fonseca should you have any questions at eduardo.fonseca@URL.", "## About FSD50K\n\nFreesound Dataset 50k (or FSD50K for short) is an open dataset of human-labeled sound events containing 51,197 <a href=\"URL clips unequally distributed in 200 classes drawn from the <a href=\"URL Ontology</a> [1]. FSD50K has been created at the <a href=\"URL Technology Group of Universitat Pompeu Fabra</a>.\n\nWhat follows is a brief summary of FSD50K's most important characteristics. Please have a look at our paper (especially Section 4) to extend the basic information provided here with relevant details for its usage, as well as discussion, limitations, applications and more.\n\n\nBasic characteristics:\n\n- FSD50K is composed mainly of sound events produced by physical sound sources and production mechanisms. \n- Following AudioSet Ontology’s main families, the FSD50K vocabulary encompasses mainly *Human sounds*, *Sounds of things*, *Animal*, *Natural sounds* and *Music*.\n- The dataset has 200 sound classes (144 leaf nodes and 56 intermediate nodes) hierarchically organized with a subset of the AudioSet Ontology. The vocabulary can be inspected in 'URL' (see Files section below).\n- FSD50K contains 51,197 audio clips totalling 108.3 hours of audio.\n- The audio content has been manually labeled by humans following a data labeling process using the <a href=\"URL Annotator</a> platform [2]. \n- Clips are of variable length from 0.3 to 30s, due to the diversity of the sound classes and the preferences of Freesound users when recording sounds.\n- Ground truth labels are provided at the clip-level (i.e., weak labels).\n- The dataset poses mainly a multi-label sound event classification problem (but also allows a variety of sound event research tasks, see Sec. 4D).\n- All clips are provided as uncompressed PCM 16 bit 44.1 kHz mono audio files.\n- The audio clips are grouped into a development (*dev*) set and an evaluation (*eval*) set such that they do not have clips from the same Freesound uploader.\n\nDev set:\n\n- 40,966 audio clips totalling 80.4 hours of audio\n- Avg duration/clip: 7.1s\n- 114,271 smeared labels (i.e., labels propagated in the upwards direction to the root of the ontology)\n- Labels are correct but could be occasionally incomplete\n- A train/validation split is provided (Sec. 3H). If a different split is used, it should be specified for reproducibility and fair comparability of results (see Sec. 5C of our paper) \n\n\nEval set:\n\n- 10,231 audio clips totalling 27.9 hours of audio\n- Avg duration/clip: 9.8s\n- 38,596 smeared labels\n- Eval set is labeled exhaustively (labels are correct and complete for the considered vocabulary)\n\n\nNOTE: All classes in FSD50K are represented in AudioSet, except 'Crash cymbal', 'Human group actions', 'Human voice', 'Respiratory sounds', and 'Domestic sounds, home sounds'.", "## License\n\nAll audio clips in FSD50K are released under Creative Commons (CC) licenses. Each clip has its own license as defined by the clip uploader in Freesound, some of them requiring attribution to their original authors and some forbidding further commercial reuse. For attribution purposes and to facilitate attribution of these files to third parties, we include a mapping from the audio clips to their corresponding licenses. The licenses are specified in the files 'dev_clips_info_FSD50K.json' and 'eval_clips_info_FSD50K.json'. These licenses are CC0, CC-BY, CC-BY-NC and CC Sampling+.\n\nIn addition, FSD50K as a whole is the result of a curation process and it has an additional license: FSD50K is released under <a href=\"URL This license is specified in the 'LICENSE-DATASET' file downloaded with the 'URL' zip file.", "## Files\n\nFSD50K can be downloaded as a series of zip files with the following directory structure:\n\n<div class=\"highlight\"><pre><span></span>root\n│ \n└───clips/ Audio clips\n│ │ \n│ └─── dev/ Audio clips in the dev set\n│ │\n│ └─── eval/ Audio clips in the eval set\n│ \n└───labels/ Files for FSD50K's ground truth\n│ │ \n│ └─── URL \t\t\t\t Ground truth for the dev set\n│ │ \n│ └─── URL \t\t\t\t Ground truth for the eval set \n│ │ \n│ └─── URL List of 200 sound classes in FSD50K \n│ \n└───metadata/ Files for additional metadata\n│ │ \n│ └─── class_info_FSD50K.json Metadata about the sound classes\n│ │ \n│ └─── dev_clips_info_FSD50K.json Metadata about the dev clips\n│ │ \n│ └─── eval_clips_info_FSD50K.json Metadata about the eval clips\n│ │ \n│ └─── pp_pnp_ratings_FSD50K.json PP/PNP ratings \n│ │ \n│ └─── collection/ Files for the *sound collection* format \n│ \n│ \n└───URL The dataset description file that you are reading\n│ \n└───LICENSE-DATASET License of the FSD50K dataset as an entity \n</pre></div>\n\n\nEach row (i.e. audio clip) of 'URL' contains the following information:\n\n- 'fname': the file name without the '.wav' extension, e.g., the fname '64760' corresponds to the file 'URL' in disk. This number is the Freesound id. We always use Freesound ids as filenames.\n- 'labels': the class labels (i.e., the ground truth). Note these class labels are *smeared*, i.e., the labels have been propagated in the upwards direction to the root of the ontology. More details about the label smearing process can be found in Appendix D of our paper. \n- 'mids': the Freebase identifiers corresponding to the class labels, as defined in the <a href=\"URL Ontology specification</a>\n- 'split': whether the clip belongs to *train* or *val* (see paper for details on the proposed split)\n\nRows in 'URL' follow the same format, except that there is no 'split' column.\n\nNOTE: We use a slightly different format than AudioSet for the naming of class labels in order to avoid potential problems with spaces, commas, etc. Example: we use 'Accelerating_and_revving_and_vroom' instead of the original 'Accelerating, revving, vroom'. You can go back to the original AudioSet naming using the information provided in 'URL' (class label and mid for the 200 classes of FSD50K) and the <a href=\"URL Ontology specification</a>.", "### Files with additional metadata (metadata/)\n\nTo allow a variety of analysis and approaches with FSD50K, we provide the following metadata:\n\n 1. 'class_info_FSD50K.json': python dictionary where each entry corresponds to one sound class and contains: 'FAQs' utilized during the annotation of the class, 'examples' (representative audio clips), and 'verification_examples' (audio clips presented to raters during annotation as a quality control mechanism). Audio clips are described by the Freesound id.\n NOTE: It may be that some of these examples are not included in the FSD50K release.\n \n 2. 'dev_clips_info_FSD50K.json': python dictionary where each entry corresponds to one dev clip and contains: title, description, tags, clip license, and the uploader name. All these metadata are provided by the uploader.\n\n 3. 'eval_clips_info_FSD50K.json': same as before, but with eval clips.\n \n 4. 'pp_pnp_ratings.json': python dictionary where each entry corresponds to one clip in the dataset and contains the PP/PNP ratings for the labels associated with the clip. More specifically, these ratings are gathered for the labels validated in the validation task (Sec. 3 of paper). This file includes 59,485 labels for the 51,197 clips in FSD50K. Out of these labels:\n\n - 56,095 labels have inter-annotator agreement (PP twice, or PNP twice). Each of these combinations can be occasionally accompanied by other (non-positive) ratings. \n - 3390 labels feature other rating configurations such as *i)* only one PP rating and one PNP rating (and nothing else). This can be considered inter-annotator agreement at the ''Present” level; *ii)* only one PP rating (and nothing else); *iii)* only one PNP rating (and nothing else).\n\n Ratings' legend: PP=1; PNP=0.5; U=0; NP=-1.\n\n NOTE: The PP/PNP ratings have been provided in the *validation* task. Subsequently, a subset of these clips corresponding to the eval set was exhaustively labeled in the *refinement* task, hence receiving additional labels in many cases. For these eval clips, you might want to check their labels in 'URL' in order to have more info about their audio content (see Sec. 3 for details).\n \n 5. 'collection/': This folder contains metadata for what we call the *sound collection format*. This format consists of the raw annotations gathered, featuring all generated class labels without any restriction. \n\n We provide the *collection* format to make available some annotations that do not appear in the FSD50K *ground truth* release. This typically happens in the case of classes for which we gathered human-provided annotations, but that were discarded in the FSD50K release due to data scarcity (more specifically, they were merged with their parents). In other words, the main purpose of the 'collection' format is to make available annotations for tiny classes. The format of these files in analogous to that of the files in 'FSD50K.ground_truth/'. A couple of examples show the differences between collection and ground truth formats:\n \n 'clip': 'labels_in_collection' -- 'labels_in_ground_truth'\n\n '51690': 'Owl' -- 'Bird,Wild_Animal,Animal'\n\n '190579': 'Toothbrush,Electric_toothbrush' -- 'Domestic_sounds_and_home_sounds'\n\n In the first example, raters provided the label 'Owl'. However, due to data scarcity, 'Owl' labels were merged into their parent 'Bird'. Then, labels 'Wild_Animal,Animal' were added via label propagation (smearing). The second example shows one of the most extreme cases, where raters provided the labels 'Electric_toothbrush,Toothbrush', which both had few data. Hence, they were merged into Toothbrush's parent, which unfortunately is 'Domestic_sounds_and_home_sounds' (a rather vague class containing a variety of children sound classes).\n\n NOTE: Labels in the collection format are not smeared. \n NOTE: While in FSD50K's ground truth the vocabulary encompasses 200 classes (common for dev and eval), since the *collection* format is composed of raw annotations, the vocabulary here is much larger (over 350 classes), and it is slightly different in dev and eval.\n\nFor further questions, please contact eduardo.fonseca@URL, or join the <a href=\"URL Google Group</a>.", "## Download\n\nClone this repository:", "## Baseline System\n\nSeveral baseline systems for FSD50K are available at <a href=\"URL/URL The experiments are described in Sec 5 of our paper.", "## References and links\n\n[1] Jort F Gemmeke, Daniel PW Ellis, Dylan Freedman, Aren Jansen, Wade Lawrence, R Channing Moore, Manoj Plakal, and Marvin Ritter. \"Audio set: An ontology and human-labeled dataset for audio events.\" In Proceedings of the International Conference on Acoustics, Speech and Signal Processing, 2017. [<a href=\"URL\n\n[2] Eduardo Fonseca, Jordi Pons, Xavier Favory, Frederic Font, Dmitry Bogdanov, Andres Ferraro, Sergio Oramas, Alastair Porter, and Xavier Serra. \"Freesound Datasets: A Platform for the Creation of Open Audio Datasets.\" In Proceedings of the International Conference on Music Information Retrieval, 2017. [<a href=\"URL\n\n\nCompanion site for FSD50K: <a href=\"URL/URL \nFreesound Annotator: <a href=\"URL/URL \nFreesound: <a href=\"URL\">URL</a> \nEduardo Fonseca's personal website: <a href=\"URL/URL \nMore datasets collected by us: <a href=\"URL/URL", "## Acknowledgments\n\nThe authors would like to thank everyone who contributed to FSD50K with annotations, and especially Mercedes Collado, Ceren Can, Rachit Gupta, Javier Arredondo, Gary Avendano and Sara Fernandez for their commitment and perseverance. The authors would also like to thank Daniel P.W. Ellis and Manoj Plakal from Google Research for valuable discussions. This work is partially supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 688382 <a href=\"URL and two Google Faculty Research Awards <a href=\"URL and <a href=\"URL and the Maria de Maeztu Units of Excellence Programme (MDM-2015-0502)." ]

[ "TAGS\n#task_categories-audio-classification #annotations_creators-unknown #language_creators-unknown #size_categories-10K<n<100K #source_datasets-unknown #license-cc-by-4.0 #arxiv-2010.00475 #region-us \n", "# Freesound Dataset 50k (FSD50K)", "## Important\n\nThis data set is a copy from the original one located at Zenodo.", "## Dataset Description\n- Homepage: FSD50K\n- Repository: GitHub\n- Paper: FSD50K: An Open Dataset of Human-Labeled Sound Events\n- Leaderboard: Paperswithcode Leaderboard\n\nIf you use the FSD50K dataset, or part of it, please cite our paper:\n\n>Eduardo Fonseca, Xavier Favory, Jordi Pons, Frederic Font, Xavier Serra. \"FSD50K: an Open Dataset of Human-Labeled Sound Events\", arXiv 2020.", "### Data curators\n\nEduardo Fonseca, Xavier Favory, Jordi Pons, Mercedes Collado, Ceren Can, Rachit Gupta, Javier Arredondo, Gary Avendano and Sara Fernandez", "### Contact\n\nYou are welcome to contact Eduardo Fonseca should you have any questions at eduardo.fonseca@URL.", "## About FSD50K\n\nFreesound Dataset 50k (or FSD50K for short) is an open dataset of human-labeled sound events containing 51,197 <a href=\"URL clips unequally distributed in 200 classes drawn from the <a href=\"URL Ontology</a> [1]. FSD50K has been created at the <a href=\"URL Technology Group of Universitat Pompeu Fabra</a>.\n\nWhat follows is a brief summary of FSD50K's most important characteristics. Please have a look at our paper (especially Section 4) to extend the basic information provided here with relevant details for its usage, as well as discussion, limitations, applications and more.\n\n\nBasic characteristics:\n\n- FSD50K is composed mainly of sound events produced by physical sound sources and production mechanisms. \n- Following AudioSet Ontology’s main families, the FSD50K vocabulary encompasses mainly *Human sounds*, *Sounds of things*, *Animal*, *Natural sounds* and *Music*.\n- The dataset has 200 sound classes (144 leaf nodes and 56 intermediate nodes) hierarchically organized with a subset of the AudioSet Ontology. The vocabulary can be inspected in 'URL' (see Files section below).\n- FSD50K contains 51,197 audio clips totalling 108.3 hours of audio.\n- The audio content has been manually labeled by humans following a data labeling process using the <a href=\"URL Annotator</a> platform [2]. \n- Clips are of variable length from 0.3 to 30s, due to the diversity of the sound classes and the preferences of Freesound users when recording sounds.\n- Ground truth labels are provided at the clip-level (i.e., weak labels).\n- The dataset poses mainly a multi-label sound event classification problem (but also allows a variety of sound event research tasks, see Sec. 4D).\n- All clips are provided as uncompressed PCM 16 bit 44.1 kHz mono audio files.\n- The audio clips are grouped into a development (*dev*) set and an evaluation (*eval*) set such that they do not have clips from the same Freesound uploader.\n\nDev set:\n\n- 40,966 audio clips totalling 80.4 hours of audio\n- Avg duration/clip: 7.1s\n- 114,271 smeared labels (i.e., labels propagated in the upwards direction to the root of the ontology)\n- Labels are correct but could be occasionally incomplete\n- A train/validation split is provided (Sec. 3H). If a different split is used, it should be specified for reproducibility and fair comparability of results (see Sec. 5C of our paper) \n\n\nEval set:\n\n- 10,231 audio clips totalling 27.9 hours of audio\n- Avg duration/clip: 9.8s\n- 38,596 smeared labels\n- Eval set is labeled exhaustively (labels are correct and complete for the considered vocabulary)\n\n\nNOTE: All classes in FSD50K are represented in AudioSet, except 'Crash cymbal', 'Human group actions', 'Human voice', 'Respiratory sounds', and 'Domestic sounds, home sounds'.", "## License\n\nAll audio clips in FSD50K are released under Creative Commons (CC) licenses. Each clip has its own license as defined by the clip uploader in Freesound, some of them requiring attribution to their original authors and some forbidding further commercial reuse. For attribution purposes and to facilitate attribution of these files to third parties, we include a mapping from the audio clips to their corresponding licenses. The licenses are specified in the files 'dev_clips_info_FSD50K.json' and 'eval_clips_info_FSD50K.json'. These licenses are CC0, CC-BY, CC-BY-NC and CC Sampling+.\n\nIn addition, FSD50K as a whole is the result of a curation process and it has an additional license: FSD50K is released under <a href=\"URL This license is specified in the 'LICENSE-DATASET' file downloaded with the 'URL' zip file.", "## Files\n\nFSD50K can be downloaded as a series of zip files with the following directory structure:\n\n<div class=\"highlight\"><pre><span></span>root\n│ \n└───clips/ Audio clips\n│ │ \n│ └─── dev/ Audio clips in the dev set\n│ │\n│ └─── eval/ Audio clips in the eval set\n│ \n└───labels/ Files for FSD50K's ground truth\n│ │ \n│ └─── URL \t\t\t\t Ground truth for the dev set\n│ │ \n│ └─── URL \t\t\t\t Ground truth for the eval set \n│ │ \n│ └─── URL List of 200 sound classes in FSD50K \n│ \n└───metadata/ Files for additional metadata\n│ │ \n│ └─── class_info_FSD50K.json Metadata about the sound classes\n│ │ \n│ └─── dev_clips_info_FSD50K.json Metadata about the dev clips\n│ │ \n│ └─── eval_clips_info_FSD50K.json Metadata about the eval clips\n│ │ \n│ └─── pp_pnp_ratings_FSD50K.json PP/PNP ratings \n│ │ \n│ └─── collection/ Files for the *sound collection* format \n│ \n│ \n└───URL The dataset description file that you are reading\n│ \n└───LICENSE-DATASET License of the FSD50K dataset as an entity \n</pre></div>\n\n\nEach row (i.e. audio clip) of 'URL' contains the following information:\n\n- 'fname': the file name without the '.wav' extension, e.g., the fname '64760' corresponds to the file 'URL' in disk. This number is the Freesound id. We always use Freesound ids as filenames.\n- 'labels': the class labels (i.e., the ground truth). Note these class labels are *smeared*, i.e., the labels have been propagated in the upwards direction to the root of the ontology. More details about the label smearing process can be found in Appendix D of our paper. \n- 'mids': the Freebase identifiers corresponding to the class labels, as defined in the <a href=\"URL Ontology specification</a>\n- 'split': whether the clip belongs to *train* or *val* (see paper for details on the proposed split)\n\nRows in 'URL' follow the same format, except that there is no 'split' column.\n\nNOTE: We use a slightly different format than AudioSet for the naming of class labels in order to avoid potential problems with spaces, commas, etc. Example: we use 'Accelerating_and_revving_and_vroom' instead of the original 'Accelerating, revving, vroom'. You can go back to the original AudioSet naming using the information provided in 'URL' (class label and mid for the 200 classes of FSD50K) and the <a href=\"URL Ontology specification</a>.", "### Files with additional metadata (metadata/)\n\nTo allow a variety of analysis and approaches with FSD50K, we provide the following metadata:\n\n 1. 'class_info_FSD50K.json': python dictionary where each entry corresponds to one sound class and contains: 'FAQs' utilized during the annotation of the class, 'examples' (representative audio clips), and 'verification_examples' (audio clips presented to raters during annotation as a quality control mechanism). Audio clips are described by the Freesound id.\n NOTE: It may be that some of these examples are not included in the FSD50K release.\n \n 2. 'dev_clips_info_FSD50K.json': python dictionary where each entry corresponds to one dev clip and contains: title, description, tags, clip license, and the uploader name. All these metadata are provided by the uploader.\n\n 3. 'eval_clips_info_FSD50K.json': same as before, but with eval clips.\n \n 4. 'pp_pnp_ratings.json': python dictionary where each entry corresponds to one clip in the dataset and contains the PP/PNP ratings for the labels associated with the clip. More specifically, these ratings are gathered for the labels validated in the validation task (Sec. 3 of paper). This file includes 59,485 labels for the 51,197 clips in FSD50K. Out of these labels:\n\n - 56,095 labels have inter-annotator agreement (PP twice, or PNP twice). Each of these combinations can be occasionally accompanied by other (non-positive) ratings. \n - 3390 labels feature other rating configurations such as *i)* only one PP rating and one PNP rating (and nothing else). This can be considered inter-annotator agreement at the ''Present” level; *ii)* only one PP rating (and nothing else); *iii)* only one PNP rating (and nothing else).\n\n Ratings' legend: PP=1; PNP=0.5; U=0; NP=-1.\n\n NOTE: The PP/PNP ratings have been provided in the *validation* task. Subsequently, a subset of these clips corresponding to the eval set was exhaustively labeled in the *refinement* task, hence receiving additional labels in many cases. For these eval clips, you might want to check their labels in 'URL' in order to have more info about their audio content (see Sec. 3 for details).\n \n 5. 'collection/': This folder contains metadata for what we call the *sound collection format*. This format consists of the raw annotations gathered, featuring all generated class labels without any restriction. \n\n We provide the *collection* format to make available some annotations that do not appear in the FSD50K *ground truth* release. This typically happens in the case of classes for which we gathered human-provided annotations, but that were discarded in the FSD50K release due to data scarcity (more specifically, they were merged with their parents). In other words, the main purpose of the 'collection' format is to make available annotations for tiny classes. The format of these files in analogous to that of the files in 'FSD50K.ground_truth/'. A couple of examples show the differences between collection and ground truth formats:\n \n 'clip': 'labels_in_collection' -- 'labels_in_ground_truth'\n\n '51690': 'Owl' -- 'Bird,Wild_Animal,Animal'\n\n '190579': 'Toothbrush,Electric_toothbrush' -- 'Domestic_sounds_and_home_sounds'\n\n In the first example, raters provided the label 'Owl'. However, due to data scarcity, 'Owl' labels were merged into their parent 'Bird'. Then, labels 'Wild_Animal,Animal' were added via label propagation (smearing). The second example shows one of the most extreme cases, where raters provided the labels 'Electric_toothbrush,Toothbrush', which both had few data. Hence, they were merged into Toothbrush's parent, which unfortunately is 'Domestic_sounds_and_home_sounds' (a rather vague class containing a variety of children sound classes).\n\n NOTE: Labels in the collection format are not smeared. \n NOTE: While in FSD50K's ground truth the vocabulary encompasses 200 classes (common for dev and eval), since the *collection* format is composed of raw annotations, the vocabulary here is much larger (over 350 classes), and it is slightly different in dev and eval.\n\nFor further questions, please contact eduardo.fonseca@URL, or join the <a href=\"URL Google Group</a>.", "## Download\n\nClone this repository:", "## Baseline System\n\nSeveral baseline systems for FSD50K are available at <a href=\"URL/URL The experiments are described in Sec 5 of our paper.", "## References and links\n\n[1] Jort F Gemmeke, Daniel PW Ellis, Dylan Freedman, Aren Jansen, Wade Lawrence, R Channing Moore, Manoj Plakal, and Marvin Ritter. \"Audio set: An ontology and human-labeled dataset for audio events.\" In Proceedings of the International Conference on Acoustics, Speech and Signal Processing, 2017. [<a href=\"URL\n\n[2] Eduardo Fonseca, Jordi Pons, Xavier Favory, Frederic Font, Dmitry Bogdanov, Andres Ferraro, Sergio Oramas, Alastair Porter, and Xavier Serra. \"Freesound Datasets: A Platform for the Creation of Open Audio Datasets.\" In Proceedings of the International Conference on Music Information Retrieval, 2017. [<a href=\"URL\n\n\nCompanion site for FSD50K: <a href=\"URL/URL \nFreesound Annotator: <a href=\"URL/URL \nFreesound: <a href=\"URL\">URL</a> \nEduardo Fonseca's personal website: <a href=\"URL/URL \nMore datasets collected by us: <a href=\"URL/URL", "## Acknowledgments\n\nThe authors would like to thank everyone who contributed to FSD50K with annotations, and especially Mercedes Collado, Ceren Can, Rachit Gupta, Javier Arredondo, Gary Avendano and Sara Fernandez for their commitment and perseverance. The authors would also like to thank Daniel P.W. Ellis and Manoj Plakal from Google Research for valuable discussions. This work is partially supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 688382 <a href=\"URL and two Google Faculty Research Awards <a href=\"URL and <a href=\"URL and the Maria de Maeztu Units of Excellence Programme (MDM-2015-0502)." ]

[ "task_categories:image-classification", "task_ids:multi-class-image-classification", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:extended|other-nist", "language:enx", "license:mit", "region:us" ]

[]

[ "enx" ]

[ "# Dataset Card for MNIST", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: MNIST handwritten digit database by Yann LeCun, Corinna Cortes, and CJ Burges\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nThe MNIST dataset consists of 70,000 28x28 black-and-white images of handwritten digits extracted from two NIST databases. There are 60,000 images in the training dataset and 10,000 images in the validation dataset, one class per digit so a total of 10 classes, with 7,000 images (6,000 train images and 1,000 test images) per class.\nHalf of the image were drawn by Census Bureau employees and the other half by high school students (this split is evenly distributed in the training and testing sets).", "### Supported Tasks and Leaderboards\n\n- 'image-classification': The goal of this task is to classify a given image of a handwritten digit into one of 10 classes representing integer values from 0 to 9, inclusively. The leaderboard is available here.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nA data point comprises an image and its label:", "### Data Fields\n\n- 'image': A 'PIL.Image.Image' object containing the 28x28 image. Note that when accessing the image column: 'dataset[0][\"image\"]' the image file is automatically decoded. Decoding of a large number of image files might take a significant amount of time. Thus it is important to first query the sample index before the '\"image\"' column, *i.e.* 'dataset[0][\"image\"]' should always be preferred over 'dataset[\"image\"][0]'\n- 'label': an integer between 0 and 9 representing the digit.", "### Data Splits\n\nThe data is split into training and test set. All the images in the test set were drawn by different individuals than the images in the training set. The training set contains 60,000 images and the test set 10,000 images.", "## Dataset Creation", "### Curation Rationale\n\nThe MNIST database was created to provide a testbed for people wanting to try pattern recognition methods or machine learning algorithms while spending minimal efforts on preprocessing and formatting. Images of the original dataset (NIST) were in two groups, one consisting of images drawn by Census Bureau employees and one consisting of images drawn by high school students. In NIST, the training set was built by grouping all the images of the Census Bureau employees, and the test set was built by grouping the images form the high school students.\nThe goal in building MNIST was to have a training and test set following the same distributions, so the training set contains 30,000 images drawn by Census Bureau employees and 30,000 images drawn by high school students, and the test set contains 5,000 images of each group. The curators took care to make sure all the images in the test set were drawn by different individuals than the images in the training set.", "### Source Data", "#### Initial Data Collection and Normalization\n\nThe original images from NIST were size normalized to fit a 20x20 pixel box while preserving their aspect ratio. The resulting images contain grey levels (i.e., pixels don't simply have a value of black and white, but a level of greyness from 0 to 255) as a result of the anti-aliasing technique used by the normalization algorithm. The images were then centered in a 28x28 image by computing the center of mass of the pixels, and translating the image so as to position this point at the center of the 28x28 field.", "#### Who are the source language producers?\n\nHalf of the source images were drawn by Census Bureau employees, half by high school students. According to the dataset curator, the images from the first group are more easily recognizable.", "### Annotations", "#### Annotation process\n\nThe images were not annotated after their creation: the image creators annotated their images with the corresponding label after drawing them.", "#### Who are the annotators?\n\nSame as the source data creators.", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nChris Burges, Corinna Cortes and Yann LeCun", "### Licensing Information\n\nMIT Licence", "### Contributions\n\nThanks to @sgugger for adding this dataset." ]

[ "TAGS\n#task_categories-image-classification #task_ids-multi-class-image-classification #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-extended|other-nist #language-Enxet #license-mit #region-us \n", "# Dataset Card for MNIST", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: MNIST handwritten digit database by Yann LeCun, Corinna Cortes, and CJ Burges\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nThe MNIST dataset consists of 70,000 28x28 black-and-white images of handwritten digits extracted from two NIST databases. There are 60,000 images in the training dataset and 10,000 images in the validation dataset, one class per digit so a total of 10 classes, with 7,000 images (6,000 train images and 1,000 test images) per class.\nHalf of the image were drawn by Census Bureau employees and the other half by high school students (this split is evenly distributed in the training and testing sets).", "### Supported Tasks and Leaderboards\n\n- 'image-classification': The goal of this task is to classify a given image of a handwritten digit into one of 10 classes representing integer values from 0 to 9, inclusively. The leaderboard is available here.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nA data point comprises an image and its label:", "### Data Fields\n\n- 'image': A 'PIL.Image.Image' object containing the 28x28 image. Note that when accessing the image column: 'dataset[0][\"image\"]' the image file is automatically decoded. Decoding of a large number of image files might take a significant amount of time. Thus it is important to first query the sample index before the '\"image\"' column, *i.e.* 'dataset[0][\"image\"]' should always be preferred over 'dataset[\"image\"][0]'\n- 'label': an integer between 0 and 9 representing the digit.", "### Data Splits\n\nThe data is split into training and test set. All the images in the test set were drawn by different individuals than the images in the training set. The training set contains 60,000 images and the test set 10,000 images.", "## Dataset Creation", "### Curation Rationale\n\nThe MNIST database was created to provide a testbed for people wanting to try pattern recognition methods or machine learning algorithms while spending minimal efforts on preprocessing and formatting. Images of the original dataset (NIST) were in two groups, one consisting of images drawn by Census Bureau employees and one consisting of images drawn by high school students. In NIST, the training set was built by grouping all the images of the Census Bureau employees, and the test set was built by grouping the images form the high school students.\nThe goal in building MNIST was to have a training and test set following the same distributions, so the training set contains 30,000 images drawn by Census Bureau employees and 30,000 images drawn by high school students, and the test set contains 5,000 images of each group. The curators took care to make sure all the images in the test set were drawn by different individuals than the images in the training set.", "### Source Data", "#### Initial Data Collection and Normalization\n\nThe original images from NIST were size normalized to fit a 20x20 pixel box while preserving their aspect ratio. The resulting images contain grey levels (i.e., pixels don't simply have a value of black and white, but a level of greyness from 0 to 255) as a result of the anti-aliasing technique used by the normalization algorithm. The images were then centered in a 28x28 image by computing the center of mass of the pixels, and translating the image so as to position this point at the center of the 28x28 field.", "#### Who are the source language producers?\n\nHalf of the source images were drawn by Census Bureau employees, half by high school students. According to the dataset curator, the images from the first group are more easily recognizable.", "### Annotations", "#### Annotation process\n\nThe images were not annotated after their creation: the image creators annotated their images with the corresponding label after drawing them.", "#### Who are the annotators?\n\nSame as the source data creators.", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nChris Burges, Corinna Cortes and Yann LeCun", "### Licensing Information\n\nMIT Licence", "### Contributions\n\nThanks to @sgugger for adding this dataset." ]

[ "task_categories:text-generation", "annotations_creators:unknown", "language_creators:unknown", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:cc-by-4.0", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#task_categories-text-generation #annotations_creators-unknown #language_creators-unknown #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-cc-by-4.0 #region-us \n" ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nThis is a dataset made up of two Bible translations-- NET and KJV.", "### Supported Tasks and Leaderboards", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale\n\nThe original intention is to use the dataset to \"translate\" between modern and 17th-century English. By doing so, we can potentially read understand things from that period more clearly.", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\nBefore the 18th and 19th centuries, English spelling was inconsistent. Because of this, the model often does not recognize spellings different from those in the KJV.\nThe model was trained on a relatively small amount of data, so it will not be as accurate as a model trained on a larger data set.", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "TAGS\n#region-us \n", "# Dataset Card for", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nThis is a dataset made up of two Bible translations-- NET and KJV.", "### Supported Tasks and Leaderboards", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale\n\nThe original intention is to use the dataset to \"translate\" between modern and 17th-century English. By doing so, we can potentially read understand things from that period more clearly.", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\nBefore the 18th and 19th centuries, English spelling was inconsistent. Because of this, the model often does not recognize spellings different from those in the KJV.\nThe model was trained on a relatively small amount of data, so it will not be as accurate as a model trained on a larger data set.", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "license:cc-by-4.0", "region:us" ]

[]

[]

[ "# Aurora SDGs Dataset\nThis data set contains metdata of 1.4 million research papers. The abstract plus the SDG labels for the Goals and Targets." ]

[ "TAGS\n#license-cc-by-4.0 #region-us \n", "# Aurora SDGs Dataset\nThis data set contains metdata of 1.4 million research papers. The abstract plus the SDG labels for the Goals and Targets." ]

[ "task_categories:automatic-speech-recognition", "annotations_creators:expert-generated", "language_creators:crowdsourced", "multilinguality:multilingual", "size_categories:10K<n<100K", "source_datasets:original", "language:en", "language:zh", "license:cc-by-sa-4.0", "arxiv:2112.06223", "region:us" ]

[ "2112.06223" ]

[ "en", "zh" ]

[ "# Dataset Card for ASCEND", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: URL\n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nASCEND (A Spontaneous Chinese-English Dataset) introduces a high-quality resource of spontaneous multi-turn conversational dialogue Chinese-English code-switching corpus collected in Hong Kong. ASCEND consists of 10.62 hours of spontaneous speech with a total of ~12.3K utterances. The corpus is split into 3 sets: training, validation, and test with a ratio of 8:1:1 while maintaining a balanced gender proportion on each set.", "### Supported Tasks and Leaderboards", "### Languages\n\nChinese and English", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "TAGS\n#task_categories-automatic-speech-recognition #annotations_creators-expert-generated #language_creators-crowdsourced #multilinguality-multilingual #size_categories-10K<n<100K #source_datasets-original #language-English #language-Chinese #license-cc-by-sa-4.0 #arxiv-2112.06223 #region-us \n", "# Dataset Card for ASCEND", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: URL\n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nASCEND (A Spontaneous Chinese-English Dataset) introduces a high-quality resource of spontaneous multi-turn conversational dialogue Chinese-English code-switching corpus collected in Hong Kong. ASCEND consists of 10.62 hours of spontaneous speech with a total of ~12.3K utterances. The corpus is split into 3 sets: training, validation, and test with a ratio of 8:1:1 while maintaining a balanced gender proportion on each set.", "### Supported Tasks and Leaderboards", "### Languages\n\nChinese and English", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "task_categories:text-generation", "language_creators:crowdsourced", "language_creators:expert-generated", "language_creators:machine-generated", "language_creators:found", "language_creators:other", "multilinguality:multilingual", "source_datasets:original", "language:asm", "language:ben", "language:brx", "language:doi", "language:guj", "language:hin", "language:kan", "language:kas", "language:kok", "language:mai", "language:mal", "language:mar", "language:mni", "language:nep", "language:ori", "language:pan", "language:san", "language:sid", "language:tam", "language:tel", "language:urd", "license:cc", "arxiv:2205.03018", "region:us" ]

[ "2205.03018" ]

[ "asm", "ben", "brx", "doi", "guj", "hin", "kan", "kas", "kok", "mai", "mal", "mar", "mni", "nep", "ori", "pan", "san", "sid", "tam", "tel", "urd" ]

[ "### Dataset Summary\n\n\nAksharantar is the largest publicly available transliteration dataset for 20 Indic languages. The corpus has 26M Indic language-English transliteration pairs.", "### Supported Tasks and Leaderboards", "### Languages\n\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'unique\\_identifier' (string): 3-letter language code followed by a unique number in each set (Train, Test, Val).\n* 'native word' (string): A word in Indic language.\n* 'english word' (string): Transliteration of native word in English (Romanised word).\n* 'source' (string): Source of the data.\n* 'score' (num): Character level log probability of indic word given roman word by IndicXlit (model). Pairs with average threshold of the 0.35 are considered.\n\n\nFor created data sources, depending on the destination/sampling method of a pair in a language, it will be one of:\n\n\n\t+ Dakshina Dataset\n\t+ IndicCorp\n\t+ Samanantar\n\t+ Wikidata\n\t+ Existing sources\n\t+ Named Entities Indian (AK-NEI)\n\t+ Named Entities Foreign (AK-NEF)\n\t+ Data from Uniform Sampling method. (Ak-Uni)\n\t+ Data from Most Frequent words sampling method. (Ak-Freq)", "### Data Splits\n\n\n\nDataset Creation\n----------------\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "#### Who are the source language producers?", "### Annotations\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "#### Annotation process\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "#### Who are the annotators?\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThis data is released under the following licensing scheme:\n\n\n* Manually collected data: Released under CC-BY license.\n* Mined dataset (from Samanantar and IndicCorp): Released under CC0 license.\n* Existing sources: Released under CC0 license.\n\n\nCC-BY License\n\n\n<a rel=\"license\" float=\"left\" href=\"URL\n<img src=\"URL style=\"border-style: none;\" alt=\"CC-BY\" width=\"100\"/>\n\n\n\n \n\n \n\nCC0 License Statement\n\n\n<a rel=\"license\" float=\"left\" href=\"URL\n<img src=\"URL style=\"border-style: none;\" alt=\"CC0\" width=\"100\"/>\n\n\n\n \n\n \n\n* We do not own any of the text from which this data has been extracted.\n* We license the actual packaging of the mined data under the Creative Commons CC0 license (“no rights reserved”).\n* To the extent possible under law, <a rel=\"dct:publisher\" href=\"URL AI4Bharat has waived all copyright and related or neighboring rights to Aksharantar manually collected data and existing sources.\n* This work is published from: India.", "### Contributions" ]

[ "TAGS\n#task_categories-text-generation #language_creators-crowdsourced #language_creators-expert-generated #language_creators-machine-generated #language_creators-found #language_creators-other #multilinguality-multilingual #source_datasets-original #language-Assamese #language-Bengali #language-Bodo (India) #language-Dogri (macrolanguage) #language-Gujarati #language-Hindi #language-Kannada #language-Kashmiri #language-Konkani (macrolanguage) #language-Maithili #language-Malayalam #language-Marathi #language-Manipuri #language-Nepali (macrolanguage) #language-Oriya (macrolanguage) #language-Panjabi #language-Sanskrit #language-Sidamo #language-Tamil #language-Telugu #language-Urdu #license-cc #arxiv-2205.03018 #region-us \n", "### Dataset Summary\n\n\nAksharantar is the largest publicly available transliteration dataset for 20 Indic languages. The corpus has 26M Indic language-English transliteration pairs.", "### Supported Tasks and Leaderboards", "### Languages\n\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'unique\\_identifier' (string): 3-letter language code followed by a unique number in each set (Train, Test, Val).\n* 'native word' (string): A word in Indic language.\n* 'english word' (string): Transliteration of native word in English (Romanised word).\n* 'source' (string): Source of the data.\n* 'score' (num): Character level log probability of indic word given roman word by IndicXlit (model). Pairs with average threshold of the 0.35 are considered.\n\n\nFor created data sources, depending on the destination/sampling method of a pair in a language, it will be one of:\n\n\n\t+ Dakshina Dataset\n\t+ IndicCorp\n\t+ Samanantar\n\t+ Wikidata\n\t+ Existing sources\n\t+ Named Entities Indian (AK-NEI)\n\t+ Named Entities Foreign (AK-NEF)\n\t+ Data from Uniform Sampling method. (Ak-Uni)\n\t+ Data from Most Frequent words sampling method. (Ak-Freq)", "### Data Splits\n\n\n\nDataset Creation\n----------------\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "#### Who are the source language producers?", "### Annotations\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "#### Annotation process\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "#### Who are the annotators?\n\n\nInformation in the paper. Aksharantar: Towards building open transliteration tools for the next billion users", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThis data is released under the following licensing scheme:\n\n\n* Manually collected data: Released under CC-BY license.\n* Mined dataset (from Samanantar and IndicCorp): Released under CC0 license.\n* Existing sources: Released under CC0 license.\n\n\nCC-BY License\n\n\n<a rel=\"license\" float=\"left\" href=\"URL\n<img src=\"URL style=\"border-style: none;\" alt=\"CC-BY\" width=\"100\"/>\n\n\n\n \n\n \n\nCC0 License Statement\n\n\n<a rel=\"license\" float=\"left\" href=\"URL\n<img src=\"URL style=\"border-style: none;\" alt=\"CC0\" width=\"100\"/>\n\n\n\n \n\n \n\n* We do not own any of the text from which this data has been extracted.\n* We license the actual packaging of the mined data under the Creative Commons CC0 license (“no rights reserved”).\n* To the extent possible under law, <a rel=\"dct:publisher\" href=\"URL AI4Bharat has waived all copyright and related or neighboring rights to Aksharantar manually collected data and existing sources.\n* This work is published from: India.", "### Contributions" ]

[ "region:us" ]

[]

[]

[ "# Full FLIP stability dataset\n\nThe stability dataset from flip, which is based on the meltome atlas, data has those columns:\n\n\n \n- Index from the original dataset\n- Seq_id a unique sequence ID string that is concatenated from several other IDs (also Unirep)\n- Sequence The actual protein sequence as a string\n- Target the melting point temperature of the protein TM \n- Cluster center The seq_id of the cluster center protein this sequence is assigned to. Can also be its won seq_id if this sequence is a center.\n- Cluster distance The levenstein distance of the protein to its cluster center." ]

[ "TAGS\n#region-us \n", "# Full FLIP stability dataset\n\nThe stability dataset from flip, which is based on the meltome atlas, data has those columns:\n\n\n \n- Index from the original dataset\n- Seq_id a unique sequence ID string that is concatenated from several other IDs (also Unirep)\n- Sequence The actual protein sequence as a string\n- Target the melting point temperature of the protein TM \n- Cluster center The seq_id of the cluster center protein this sequence is assigned to. Can also be its won seq_id if this sequence is a center.\n- Cluster distance The levenstein distance of the protein to its cluster center." ]

[ "license:cc-by-4.0", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-cc-by-4.0 #region-us \n" ]

[ "task_categories:token-classification", "task_ids:part-of-speech", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "language:en", "license:cc-by-4.0", "region:us" ]

[]

[ "en" ]

[ "### Dataset Summary\n\n\nPart-of-speech information is basic NLP task. However, Twitter text\nis difficult to part-of-speech tag: it is noisy, with linguistic errors and idiosyncratic style.\nThis dataset contains two datasets for English PoS tagging for tweets:\n\n\n* Ritter, with train/dev/test\n* Foster, with dev/test\n\n\nSplits defined in the Derczynski paper, but the data is from Ritter and Foster.\n\n\n* Ritter: URL\n* Foster: URL", "### Supported Tasks and Leaderboards\n\n\n* Part of speech tagging on Ritter", "### Languages\n\n\nEnglish, non-region-specific. 'bcp47:en'\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.", "#### twitter-pos\n\n\n* 'id': a 'string' feature.\n* 'tokens': a 'list' of 'string' features.\n* 'pos\\_tags': a 'list' of classification labels ('int'). Full tagset with indices:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information", "### Contributions\n\n\nAuthor uploaded (@leondz)" ]

[ "TAGS\n#task_categories-token-classification #task_ids-part-of-speech #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-English #license-cc-by-4.0 #region-us \n", "### Dataset Summary\n\n\nPart-of-speech information is basic NLP task. However, Twitter text\nis difficult to part-of-speech tag: it is noisy, with linguistic errors and idiosyncratic style.\nThis dataset contains two datasets for English PoS tagging for tweets:\n\n\n* Ritter, with train/dev/test\n* Foster, with dev/test\n\n\nSplits defined in the Derczynski paper, but the data is from Ritter and Foster.\n\n\n* Ritter: URL\n* Foster: URL", "### Supported Tasks and Leaderboards\n\n\n* Part of speech tagging on Ritter", "### Languages\n\n\nEnglish, non-region-specific. 'bcp47:en'\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.", "#### twitter-pos\n\n\n* 'id': a 'string' feature.\n* 'tokens': a 'list' of 'string' features.\n* 'pos\\_tags': a 'list' of classification labels ('int'). Full tagset with indices:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information", "### Contributions\n\n\nAuthor uploaded (@leondz)" ]

[ "license:apache-2.0", "region:us" ]

[]

[]

[ "# Dataset Card for [KOR-RE-natures-and-environments]\n\nYou can find relation map, guidelines(written in Korean), short technical papers in this github repo. This work is done by as part of project for Boostcamp AI Tech supported by Naver Connect Foundation.", "### Dataset Description\n* Language: Korean\n* Task: Relation Extraction\n* Topics: Natures and Environments\n* Sources: Korean wiki", "### Main Data Fields\n* Sentences: sentences\n* Subject_entity: infos for subject entity in the sentence including words, start index, end index, type of entity\n* object_entity: infos for object entity in the sentence including words, start index, end index, type of entity\n* label : class ground truth label\n* file : name of the file" ]

[ "TAGS\n#license-apache-2.0 #region-us \n", "# Dataset Card for [KOR-RE-natures-and-environments]\n\nYou can find relation map, guidelines(written in Korean), short technical papers in this github repo. This work is done by as part of project for Boostcamp AI Tech supported by Naver Connect Foundation.", "### Dataset Description\n* Language: Korean\n* Task: Relation Extraction\n* Topics: Natures and Environments\n* Sources: Korean wiki", "### Main Data Fields\n* Sentences: sentences\n* Subject_entity: infos for subject entity in the sentence including words, start index, end index, type of entity\n* object_entity: infos for object entity in the sentence including words, start index, end index, type of entity\n* label : class ground truth label\n* file : name of the file" ]

[ "region:us" ]

[]

[]

[ "# README", "## Annotated Student Feedback\n---\nannotations_creators: []\nlanguage:\n- en\nlicense:\n- mit\n---\nThis resource contains 3000 student feedback data that have been annotated for aspect terms, opinion terms, polarities of the opinion terms towards targeted aspects, document-level opinion polarities, and sentence separations.", "### Folder Structure of the resource,\n\n\n\nEach Annotated_part_# folders contain three files. Those are in XMI, XML, and ZIP formats. \nXMI files contain the annotated student feedback data and XML files contain tagsets used for annotation.\n\nFind the code for reading data from XML and XMI files in 'code_for_read_annotated_data.py'" ]

[ "TAGS\n#region-us \n", "# README", "## Annotated Student Feedback\n---\nannotations_creators: []\nlanguage:\n- en\nlicense:\n- mit\n---\nThis resource contains 3000 student feedback data that have been annotated for aspect terms, opinion terms, polarities of the opinion terms towards targeted aspects, document-level opinion polarities, and sentence separations.", "### Folder Structure of the resource,\n\n\n\nEach Annotated_part_# folders contain three files. Those are in XMI, XML, and ZIP formats. \nXMI files contain the annotated student feedback data and XML files contain tagsets used for annotation.\n\nFind the code for reading data from XML and XMI files in 'code_for_read_annotated_data.py'" ]

[ "task_categories:fill-mask", "task_categories:text-generation", "task_ids:masked-language-modeling", "task_ids:language-modeling", "annotations_creators:no-annotation", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:ko", "license:cc0-1.0", "region:us" ]

[]

[ "ko" ]

[ "### Dataset Summary\n\n\n아카라이브 베스트 라이브 채널의 2021년 8월 16일부터 2022년 5월 6일까지의 데이터를 수집하여, 댓글만 골라낸 데이터입니다.\n\n\n커뮤니티 특성상, 민감한 데이터들도 많으므로 사용에 주의가 필요합니다.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nko\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n* Size of downloaded dataset files: 21.3 MB", "### Data Fields\n\n\n* text: 'string'", "### Data Splits" ]

[ "TAGS\n#task_categories-fill-mask #task_categories-text-generation #task_ids-masked-language-modeling #task_ids-language-modeling #annotations_creators-no-annotation #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-Korean #license-cc0-1.0 #region-us \n", "### Dataset Summary\n\n\n아카라이브 베스트 라이브 채널의 2021년 8월 16일부터 2022년 5월 6일까지의 데이터를 수집하여, 댓글만 골라낸 데이터입니다.\n\n\n커뮤니티 특성상, 민감한 데이터들도 많으므로 사용에 주의가 필요합니다.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nko\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n* Size of downloaded dataset files: 21.3 MB", "### Data Fields\n\n\n* text: 'string'", "### Data Splits" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "task_categories:token-classification", "language:id", "region:us" ]

[]

[ "id" ]

[]

[ "TAGS\n#task_categories-token-classification #language-Indonesian #region-us \n" ]

[ "language:en", "fanfiction", "region:us" ]

[]

[ "en" ]

[ "## URL\n\nCleaning up URL\n\nStarting with \"Z\" stories to get the hang of it." ]

[ "TAGS\n#language-English #fanfiction #region-us \n", "## URL\n\nCleaning up URL\n\nStarting with \"Z\" stories to get the hang of it." ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "license:wtfpl", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-wtfpl #region-us \n" ]

[ "task_categories:text-classification", "language_creators:found", "multilinguality:monolingual", "language:en", "license:cc-by-nc-sa-4.0", "arxiv:2207.00220", "region:us" ]

[ "2207.00220" ]

[ "en" ]

[ "# Dataset Card for eoir_privacy", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nThis dataset mimics privacy standards for EOIR decisions. It is meant to help learn contextual data sanitization rules to anonymize potentially sensitive contexts in crawled language data.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\n{\n \"text\" : masked paragraph, \n \"label\" : whether to use a pseudonym in filling masks\n}", "### Data Splits\n\ntrain 75%, validation 25%", "## Dataset Creation", "### Curation Rationale\n\nThis dataset mimics privacy standards for EOIR decisions. It is meant to help learn contextual data sanitization rules to anonymize potentially sensitive contexts in crawled language data.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe scrape EOIR. We then filter at the paragraph level and replace any references to respondent, applicant, or names with [MASK] tokens. We then determine if the case used a pseudonym or not.", "#### Who are the source language producers?\n\nU.S. Executive Office for Immigration Review", "### Annotations", "#### Annotation process\n\nAnnotations (i.e., pseudonymity decisions) were made by the EOIR court. We use regex to identify if a pseudonym was used to refer to the applicant/respondent.", "#### Who are the annotators?\n\nEOIR judges.", "### Personal and Sensitive Information\n\nThere may be sensitive contexts involved, the courts already make a determination as to data filtering of sensitive data, but nonetheless there may be sensitive topics discussed.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is meant to learn contextual privacy rules to help filter private/sensitive data, but itself encodes biases of the courts from which the data came. We suggest that people look beyond this data for learning more contextual privacy rules.", "### Discussion of Biases\n\nData may be biased due to its origin in U.S. immigration courts.", "### Licensing Information\n\nCC-BY-NC" ]

[ "TAGS\n#task_categories-text-classification #language_creators-found #multilinguality-monolingual #language-English #license-cc-by-nc-sa-4.0 #arxiv-2207.00220 #region-us \n", "# Dataset Card for eoir_privacy", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nThis dataset mimics privacy standards for EOIR decisions. It is meant to help learn contextual data sanitization rules to anonymize potentially sensitive contexts in crawled language data.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\n{\n \"text\" : masked paragraph, \n \"label\" : whether to use a pseudonym in filling masks\n}", "### Data Splits\n\ntrain 75%, validation 25%", "## Dataset Creation", "### Curation Rationale\n\nThis dataset mimics privacy standards for EOIR decisions. It is meant to help learn contextual data sanitization rules to anonymize potentially sensitive contexts in crawled language data.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe scrape EOIR. We then filter at the paragraph level and replace any references to respondent, applicant, or names with [MASK] tokens. We then determine if the case used a pseudonym or not.", "#### Who are the source language producers?\n\nU.S. Executive Office for Immigration Review", "### Annotations", "#### Annotation process\n\nAnnotations (i.e., pseudonymity decisions) were made by the EOIR court. We use regex to identify if a pseudonym was used to refer to the applicant/respondent.", "#### Who are the annotators?\n\nEOIR judges.", "### Personal and Sensitive Information\n\nThere may be sensitive contexts involved, the courts already make a determination as to data filtering of sensitive data, but nonetheless there may be sensitive topics discussed.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is meant to learn contextual privacy rules to help filter private/sensitive data, but itself encodes biases of the courts from which the data came. We suggest that people look beyond this data for learning more contextual privacy rules.", "### Discussion of Biases\n\nData may be biased due to its origin in U.S. immigration courts.", "### Licensing Information\n\nCC-BY-NC" ]

[ "task_categories:text-classification", "task_ids:fact-checking", "task_ids:sentiment-classification", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:n<1K", "source_datasets:original", "language:ru", "license:cc-by-4.0", "stance-detection", "arxiv:1809.01574", "region:us" ]

[ "1809.01574" ]

[ "ru" ]

[ "### Dataset Summary\n\n\nThis is a stance prediction dataset in Russian. The dataset contains comments on news articles,\nand rows are a comment, the title of the news article it responds to, and the stance of the comment\ntowards the article.\n\n\nStance detection is a critical component of rumour and fake news identification. It involves the extraction of the stance a particular author takes related to a given claim, both expressed in text. This paper investigates stance classification for Russian. It introduces a new dataset, RuStance, of Russian tweets and news comments from multiple sources, covering multiple stories, as well as text classification approaches to stance detection as benchmarks over this data in this language. As well as presenting this openly-available dataset, the first of its kind for Russian, the paper presents a baseline for stance prediction in the language.", "### Supported Tasks and Leaderboards\n\n\n* Stance Detection: Stance Detection on RuStance", "### Languages\n\n\nRussian, as spoken on the Meduza website (i.e. from multiple countries) ('bcp47:ru')\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### rustance\n\n\n* Size of downloaded dataset files: 349.79 KiB\n* Size of the generated dataset: 366.11 KiB\n* Total amount of disk used: 715.90 KiB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature.\n* 'text': a 'string' expressing a stance.\n* 'title': a 'string' of the target/topic annotated here.\n* 'stance': a class label representing the stance the text expresses towards the target. Full tagset with indices:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nToy data for training and especially evaluating stance prediction in Russian", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe data is comments scraped from a Russian news site not situated in Russia, Meduza, in 2018.", "#### Who are the source language producers?\n\n\nRussian speakers including from the Russian diaspora, especially Latvia", "### Annotations", "#### Annotation process\n\n\nAnnotators labelled comments for supporting, denying, querying or just commenting on a news article.", "#### Who are the annotators?\n\n\nRussian native speakers, IT education, male, 20s.", "### Personal and Sensitive Information\n\n\nThe data was public at the time of collection. No PII removal has been performed.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThere's a risk of misinformative content being in this data. The data has NOT been vetted for any content.", "### Discussion of Biases", "### Other Known Limitations\n\n\nThe above limitations apply.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors.", "### Licensing Information\n\n\nThe authors distribute this data under Creative Commons attribution license, CC-BY 4.0.", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "TAGS\n#task_categories-text-classification #task_ids-fact-checking #task_ids-sentiment-classification #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-n<1K #source_datasets-original #language-Russian #license-cc-by-4.0 #stance-detection #arxiv-1809.01574 #region-us \n", "### Dataset Summary\n\n\nThis is a stance prediction dataset in Russian. The dataset contains comments on news articles,\nand rows are a comment, the title of the news article it responds to, and the stance of the comment\ntowards the article.\n\n\nStance detection is a critical component of rumour and fake news identification. It involves the extraction of the stance a particular author takes related to a given claim, both expressed in text. This paper investigates stance classification for Russian. It introduces a new dataset, RuStance, of Russian tweets and news comments from multiple sources, covering multiple stories, as well as text classification approaches to stance detection as benchmarks over this data in this language. As well as presenting this openly-available dataset, the first of its kind for Russian, the paper presents a baseline for stance prediction in the language.", "### Supported Tasks and Leaderboards\n\n\n* Stance Detection: Stance Detection on RuStance", "### Languages\n\n\nRussian, as spoken on the Meduza website (i.e. from multiple countries) ('bcp47:ru')\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### rustance\n\n\n* Size of downloaded dataset files: 349.79 KiB\n* Size of the generated dataset: 366.11 KiB\n* Total amount of disk used: 715.90 KiB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature.\n* 'text': a 'string' expressing a stance.\n* 'title': a 'string' of the target/topic annotated here.\n* 'stance': a class label representing the stance the text expresses towards the target. Full tagset with indices:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nToy data for training and especially evaluating stance prediction in Russian", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe data is comments scraped from a Russian news site not situated in Russia, Meduza, in 2018.", "#### Who are the source language producers?\n\n\nRussian speakers including from the Russian diaspora, especially Latvia", "### Annotations", "#### Annotation process\n\n\nAnnotators labelled comments for supporting, denying, querying or just commenting on a news article.", "#### Who are the annotators?\n\n\nRussian native speakers, IT education, male, 20s.", "### Personal and Sensitive Information\n\n\nThe data was public at the time of collection. No PII removal has been performed.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThere's a risk of misinformative content being in this data. The data has NOT been vetted for any content.", "### Discussion of Biases", "### Other Known Limitations\n\n\nThe above limitations apply.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors.", "### Licensing Information\n\n\nThe authors distribute this data under Creative Commons attribution license, CC-BY 4.0.", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "task_categories:audio-classification", "annotations_creators:unknown", "language_creators:unknown", "size_categories:100K<n<100M", "source_datasets:unknown", "license:cc-by-sa-4.0", "audio-slot-filling", "region:us" ]

[]

[]

[ "# Dataset Card for audioset2022", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: AudioSet Ontology\n- Repository: \n- Paper: Audio Set: An ontology and human-labeled dataset for audio events\n- Leaderboard: Paperswithcode Leaderboard", "### Dataset Summary\n\nThe AudioSet ontology is a collection of sound events organized in a hierarchy. The ontology covers a wide range of everyday sounds, from human and animal sounds, to natural and environmental sounds, to musical and miscellaneous sounds.\n\nThis repository only includes audio files for DCASE 2022 - Task 3\n\nThe included labels are limited to:\n- Female speech, woman speaking\n- Male speech, man speaking\n- Clapping\n- Telephone\n- Telephone bell ringing\n- Ringtone\n- Laughter\n- Domestic sounds, home sounds\n- Vacuum cleaner\n- Kettle whistle\n- Mechanical fan\n- Walk, footsteps\n- Door\n- Cupboard open or close\n- Music\n- Background music\n- Pop music\n- Musical instrument\n- Acoustic guitar\n- Marimba, xylophone\n- Cowbell\n- Piano\n- Electric piano\n- Rattle (instrument)\n- Water tap, faucet\n- Bell\n- Bicycle bell\n- Chime\n- Knock", "### Supported Tasks and Leaderboards\n\n- 'audio-classification': The dataset can be used to train a model for Sound Event Detection/Localization.\n\nThe recordings only includes the single channel audio. For Localization tasks, it will required to apply RIR information", "### Languages\n\nNone", "## Dataset Structure", "### Data Instances\n\nWIP", "### Data Fields\n\n- file: A path to the downloaded audio file in .mp3 format.", "### Data Splits\n\nThis dataset only includes audio file from the unbalance train list.\nThe data comprises two splits: weak labels and strong labels.", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\nThe dataset was initially downloaded by Nelson Yalta (URL@URL).", "### Licensing Information\nCC BY-SA 4.0" ]

[ "TAGS\n#task_categories-audio-classification #annotations_creators-unknown #language_creators-unknown #size_categories-100K<n<100M #source_datasets-unknown #license-cc-by-sa-4.0 #audio-slot-filling #region-us \n", "# Dataset Card for audioset2022", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: AudioSet Ontology\n- Repository: \n- Paper: Audio Set: An ontology and human-labeled dataset for audio events\n- Leaderboard: Paperswithcode Leaderboard", "### Dataset Summary\n\nThe AudioSet ontology is a collection of sound events organized in a hierarchy. The ontology covers a wide range of everyday sounds, from human and animal sounds, to natural and environmental sounds, to musical and miscellaneous sounds.\n\nThis repository only includes audio files for DCASE 2022 - Task 3\n\nThe included labels are limited to:\n- Female speech, woman speaking\n- Male speech, man speaking\n- Clapping\n- Telephone\n- Telephone bell ringing\n- Ringtone\n- Laughter\n- Domestic sounds, home sounds\n- Vacuum cleaner\n- Kettle whistle\n- Mechanical fan\n- Walk, footsteps\n- Door\n- Cupboard open or close\n- Music\n- Background music\n- Pop music\n- Musical instrument\n- Acoustic guitar\n- Marimba, xylophone\n- Cowbell\n- Piano\n- Electric piano\n- Rattle (instrument)\n- Water tap, faucet\n- Bell\n- Bicycle bell\n- Chime\n- Knock", "### Supported Tasks and Leaderboards\n\n- 'audio-classification': The dataset can be used to train a model for Sound Event Detection/Localization.\n\nThe recordings only includes the single channel audio. For Localization tasks, it will required to apply RIR information", "### Languages\n\nNone", "## Dataset Structure", "### Data Instances\n\nWIP", "### Data Fields\n\n- file: A path to the downloaded audio file in .mp3 format.", "### Data Splits\n\nThis dataset only includes audio file from the unbalance train list.\nThe data comprises two splits: weak labels and strong labels.", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\nThe dataset was initially downloaded by Nelson Yalta (URL@URL).", "### Licensing Information\nCC BY-SA 4.0" ]

[ "task_categories:summarization", "task_categories:text2text-generation", "multilinguality:monolingual", "size_categories:1K<n<10K", "language:en", "conditional-text-generation", "arxiv:2005.10070", "arxiv:2110.08499", "region:us" ]

[ "2005.10070", "2110.08499" ]

[ "en" ]

[ "### Data Fields\n\n\n* 'id': paper id\n* 'document': a string/list containing the body of a set of documents\n* 'summary': a string containing the abstract of the set", "### Data Splits\n\n\nThis dataset has 3 splits: *train*, *validation*, and *test*. \\\n\n\n\nCite original article\n=====================" ]

[ "TAGS\n#task_categories-summarization #task_categories-text2text-generation #multilinguality-monolingual #size_categories-1K<n<10K #language-English #conditional-text-generation #arxiv-2005.10070 #arxiv-2110.08499 #region-us \n", "### Data Fields\n\n\n* 'id': paper id\n* 'document': a string/list containing the body of a set of documents\n* 'summary': a string containing the abstract of the set", "### Data Splits\n\n\nThis dataset has 3 splits: *train*, *validation*, and *test*. \\\n\n\n\nCite original article\n=====================" ]

[ "license:afl-3.0", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-afl-3.0 #region-us \n" ]

[ "license:other", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-other #region-us \n" ]

[ "license:mit", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-mit #region-us \n" ]

[ "task_categories:text2text-generation", "task_ids:text-simplification", "annotations_creators:machine-generated", "annotations_creators:crowdsourced", "annotations_creators:found", "language_creators:machine-generated", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "source_datasets:extended|squad", "source_datasets:extended|race", "source_datasets:extended|newsqa", "source_datasets:extended|qamr", "source_datasets:extended|movieQA", "license:mit", "arxiv:1809.02922", "region:us" ]

[ "1809.02922" ]

[]

[ "### Dataset Summary\n\n\nExisting datasets for natural language inference (NLI) have propelled research on language understanding. We propose a new method for automatically deriving NLI datasets from the growing abundance of large-scale question answering datasets. Our approach hinges on learning a sentence transformation model which converts question-answer pairs into their declarative forms. Despite being primarily trained on a single QA dataset, we show that it can be successfully applied to a variety of other QA resources. Using this system, we automatically derive a new freely available dataset of over 500k NLI examples (QA-NLI), and show that it exhibits a wide range of inference phenomena rarely seen in previous NLI datasets.\n\n\nThis Question to Declarative Sentence (QA2D) Dataset contains 86k question-answer pairs and their manual transformation into declarative sentences. 95% of question answer pairs come from SQuAD (Rajkupar et al., 2016) and the remaining 5% come from four other question answering datasets.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nen\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nSee below.", "### Data Fields\n\n\n* 'dataset': lowercased name of dataset (movieqa, newsqa, qamr, race, squad)\n* 'example\\_uid': unique id of example within dataset (there are examples with the same uids from different datasets, so the combination of dataset + example\\_uid should be used for unique indexing)\n* 'question': tokenized (space-separated) question from the source QA dataset\n* 'answer': tokenized (space-separated) answer span from the source QA dataset\n* 'turker\\_answer': tokenized (space-separated) answer sentence collected from MTurk\n* 'rule-based': tokenized (space-separated) answer sentence, generated by the rule-based model", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nThis Question to Declarative Sentence (QA2D) Dataset contains 86k question-answer pairs and their manual transformation into declarative sentences. 95% of question answer pairs come from SQuAD (Rajkupar et al., 2016) and the remaining 5% come from four other question answering datasets.", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\n@article{DBLP:journals/corr/abs-1809-02922,\nauthor = {Dorottya Demszky and\nKelvin Guu and\nPercy Liang},\ntitle = {Transforming Question Answering Datasets Into Natural Language Inference\nDatasets},\njournal = {CoRR},\nvolume = {abs/1809.02922},\nyear = {2018},\nurl = {URL\neprinttype = {arXiv},\neprint = {1809.02922},\ntimestamp = {Fri, 05 Oct 2018 11:34:52 +0200},\nbiburl = {URL\nbibsource = {dblp computer science bibliography, URL}\n}" ]

[ "TAGS\n#task_categories-text2text-generation #task_ids-text-simplification #annotations_creators-machine-generated #annotations_creators-crowdsourced #annotations_creators-found #language_creators-machine-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #source_datasets-extended|squad #source_datasets-extended|race #source_datasets-extended|newsqa #source_datasets-extended|qamr #source_datasets-extended|movieQA #license-mit #arxiv-1809.02922 #region-us \n", "### Dataset Summary\n\n\nExisting datasets for natural language inference (NLI) have propelled research on language understanding. We propose a new method for automatically deriving NLI datasets from the growing abundance of large-scale question answering datasets. Our approach hinges on learning a sentence transformation model which converts question-answer pairs into their declarative forms. Despite being primarily trained on a single QA dataset, we show that it can be successfully applied to a variety of other QA resources. Using this system, we automatically derive a new freely available dataset of over 500k NLI examples (QA-NLI), and show that it exhibits a wide range of inference phenomena rarely seen in previous NLI datasets.\n\n\nThis Question to Declarative Sentence (QA2D) Dataset contains 86k question-answer pairs and their manual transformation into declarative sentences. 95% of question answer pairs come from SQuAD (Rajkupar et al., 2016) and the remaining 5% come from four other question answering datasets.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nen\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nSee below.", "### Data Fields\n\n\n* 'dataset': lowercased name of dataset (movieqa, newsqa, qamr, race, squad)\n* 'example\\_uid': unique id of example within dataset (there are examples with the same uids from different datasets, so the combination of dataset + example\\_uid should be used for unique indexing)\n* 'question': tokenized (space-separated) question from the source QA dataset\n* 'answer': tokenized (space-separated) answer span from the source QA dataset\n* 'turker\\_answer': tokenized (space-separated) answer sentence collected from MTurk\n* 'rule-based': tokenized (space-separated) answer sentence, generated by the rule-based model", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nThis Question to Declarative Sentence (QA2D) Dataset contains 86k question-answer pairs and their manual transformation into declarative sentences. 95% of question answer pairs come from SQuAD (Rajkupar et al., 2016) and the remaining 5% come from four other question answering datasets.", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\n@article{DBLP:journals/corr/abs-1809-02922,\nauthor = {Dorottya Demszky and\nKelvin Guu and\nPercy Liang},\ntitle = {Transforming Question Answering Datasets Into Natural Language Inference\nDatasets},\njournal = {CoRR},\nvolume = {abs/1809.02922},\nyear = {2018},\nurl = {URL\neprinttype = {arXiv},\neprint = {1809.02922},\ntimestamp = {Fri, 05 Oct 2018 11:34:52 +0200},\nbiburl = {URL\nbibsource = {dblp computer science bibliography, URL}\n}" ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for github_issues_300", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: URL\n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nGitHub issues dataset as in the Hugging Face course (URL but restricted to 300 issues", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "TAGS\n#region-us \n", "# Dataset Card for github_issues_300", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: URL\n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nGitHub issues dataset as in the Hugging Face course (URL but restricted to 300 issues", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "task_categories:token-classification", "task_ids:parsing", "annotations_creators:expert-generated", "language_creators:found", "region:us" ]

[]

[]

[ "# Information Card for Brat", "## Table of Contents\n\n- Description\n - Summary\n- Dataset Structure\n- Data Instances\n- Data Fields\n- Usage\n- Additional Information\n - Licensing Information\n - Citation Information", "## Description\n\n- Homepage: URL\n- Paper: URL\n- Leaderboard: \\\\]\n- Point of Contact: \\\\]", "### Summary\n\nBrat is an intuitive web-based tool for text annotation supported by Natural Language Processing (NLP) technology. BRAT has been developed for rich structured annota- tion for a variety of NLP tasks and aims to support manual curation efforts and increase annotator productivity using NLP techniques. brat is designed in particular for structured annotation, where the notes are not free form text but have a fixed form that can be automatically processed and interpreted by a computer.", "## Dataset Structure\n\nDataset annotated with brat format is processed using this script. Annotations created in brat are stored on disk in a standoff format: annotations are stored separately from the annotated document text, which is never modified by the tool. For each text document in the system, there is a corresponding annotation file. The two are associated by the file naming convention that their base name (file name without suffix) is the same: for example, the file URL contains annotations for the file URL. More information can be found here.", "### Data Instances", "### Data Fields\n\n- 'context' ('str'): the textual content of the data file\n- 'file_name' ('str'): the name of the data / annotation file without extension\n- 'spans' ('dict'): span annotations of the 'context' string\n - 'id' ('str'): the id of the span, starts with 'T'\n - 'type' ('str'): the label of the span\n - 'locations' ('list'): the indices indicating the span's locations (multiple because of fragments), consisting of 'dict's with\n - 'start' ('list' of 'int'): the indices indicating the inclusive character start positions of the span fragments\n - 'end' ('list' of 'int'): the indices indicating the exclusive character end positions of the span fragments\n - 'text' ('list' of 'str'): the texts of the span fragments\n- 'relations': a sequence of relations between elements of 'spans'\n - 'id' ('str'): the id of the relation, starts with 'R'\n - 'type' ('str'): the label of the relation\n - 'arguments' ('list' of 'dict'): the spans related to the relation, consisting of 'dict's with\n - 'type' ('list' of 'str'): the argument roles of the spans in the relation, either 'Arg1' or 'Arg2'\n - 'target' ('list' of 'str'): the spans which are the arguments of the relation\n- 'equivalence_relations': contains 'type' and 'target' (more information needed)\n- 'events': contains 'id', 'type', 'trigger', and 'arguments' (more information needed)\n- 'attributions' ('dict'): attribute annotations of any other annotation\n - 'id' ('str'): the instance id of the attribution\n - 'type' ('str'): the type of the attribution\n - 'target' ('str'): the id of the annotation to which the attribution is for\n - 'value' ('str'): the attribution's value or mark\n- 'normalizations' ('dict'): the unique identification of the real-world entities referred to by specific text expressions\n - 'id' ('str'): the instance id of the normalized entity\n - 'type'('str'): the type of the normalized entity\n - 'target' ('str'): the id of the annotation to which the normalized entity is for\n - 'resource_id' ('str'): the associated resource to the normalized entity\n - 'entity_id' ('str'): the instance id of normalized entity\n- 'notes' ('dict'): a freeform text, added to the annotation\n - 'id' ('str'): the instance id of the note\n - 'type' ('str'): the type of note\n - 'target' ('str'): the id of the related annotation\n - 'note' ('str'): the text body of the note", "### Usage\n\nThe 'brat' dataset script can be used by calling 'load_dataset()' method and passing any arguments that are accepted by the 'BratConfig' (which is a special BuilderConfig). It requires at least the 'url' argument. The full list of arguments is as follows:\n\n- 'url' ('str'): the url of the dataset which should point to either a zip file or a directory containing the Brat data ('*.txt') and annotation ('*.ann') files\n\n- 'description' ('str', optional): the description of the dataset\n\n- 'citation' ('str', optional): the citation of the dataset\n\n- 'homepage' ('str', optional): the homepage of the dataset\n\n- 'split_paths' ('dict', optional): a mapping of (arbitrary) split names to subdirectories or lists of files (without extension), e.g. '{\"train\": \"path/to/train_directory\", \"test\": \"path/to/test_director\"}' or '{\"train\": [\"path/to/train_file1\", \"path/to/train_file2\"]}'. In both cases (subdirectory paths or file paths), the paths are relative to the url. If 'split_paths' is not provided, the dataset will be loaded from the root directory and all direct subfolders will be considered as splits.\n\n- 'file_name_blacklist' ('list', optional): a list of file names (without extension) that should be ignored, e.g. '[\"A28\"]'. This is useful if the dataset contains files that are not valid brat files.\n\nImportant: Using the 'data_dir' parameter of the 'load_dataset()' method overrides the 'url' parameter of the 'BratConfig'.\n\nWe provide an example of SciArg dataset below:", "## Additional Information", "### Licensing Information\n\n\\\\]" ]

[ "TAGS\n#task_categories-token-classification #task_ids-parsing #annotations_creators-expert-generated #language_creators-found #region-us \n", "# Information Card for Brat", "## Table of Contents\n\n- Description\n - Summary\n- Dataset Structure\n- Data Instances\n- Data Fields\n- Usage\n- Additional Information\n - Licensing Information\n - Citation Information", "## Description\n\n- Homepage: URL\n- Paper: URL\n- Leaderboard: \\\\]\n- Point of Contact: \\\\]", "### Summary\n\nBrat is an intuitive web-based tool for text annotation supported by Natural Language Processing (NLP) technology. BRAT has been developed for rich structured annota- tion for a variety of NLP tasks and aims to support manual curation efforts and increase annotator productivity using NLP techniques. brat is designed in particular for structured annotation, where the notes are not free form text but have a fixed form that can be automatically processed and interpreted by a computer.", "## Dataset Structure\n\nDataset annotated with brat format is processed using this script. Annotations created in brat are stored on disk in a standoff format: annotations are stored separately from the annotated document text, which is never modified by the tool. For each text document in the system, there is a corresponding annotation file. The two are associated by the file naming convention that their base name (file name without suffix) is the same: for example, the file URL contains annotations for the file URL. More information can be found here.", "### Data Instances", "### Data Fields\n\n- 'context' ('str'): the textual content of the data file\n- 'file_name' ('str'): the name of the data / annotation file without extension\n- 'spans' ('dict'): span annotations of the 'context' string\n - 'id' ('str'): the id of the span, starts with 'T'\n - 'type' ('str'): the label of the span\n - 'locations' ('list'): the indices indicating the span's locations (multiple because of fragments), consisting of 'dict's with\n - 'start' ('list' of 'int'): the indices indicating the inclusive character start positions of the span fragments\n - 'end' ('list' of 'int'): the indices indicating the exclusive character end positions of the span fragments\n - 'text' ('list' of 'str'): the texts of the span fragments\n- 'relations': a sequence of relations between elements of 'spans'\n - 'id' ('str'): the id of the relation, starts with 'R'\n - 'type' ('str'): the label of the relation\n - 'arguments' ('list' of 'dict'): the spans related to the relation, consisting of 'dict's with\n - 'type' ('list' of 'str'): the argument roles of the spans in the relation, either 'Arg1' or 'Arg2'\n - 'target' ('list' of 'str'): the spans which are the arguments of the relation\n- 'equivalence_relations': contains 'type' and 'target' (more information needed)\n- 'events': contains 'id', 'type', 'trigger', and 'arguments' (more information needed)\n- 'attributions' ('dict'): attribute annotations of any other annotation\n - 'id' ('str'): the instance id of the attribution\n - 'type' ('str'): the type of the attribution\n - 'target' ('str'): the id of the annotation to which the attribution is for\n - 'value' ('str'): the attribution's value or mark\n- 'normalizations' ('dict'): the unique identification of the real-world entities referred to by specific text expressions\n - 'id' ('str'): the instance id of the normalized entity\n - 'type'('str'): the type of the normalized entity\n - 'target' ('str'): the id of the annotation to which the normalized entity is for\n - 'resource_id' ('str'): the associated resource to the normalized entity\n - 'entity_id' ('str'): the instance id of normalized entity\n- 'notes' ('dict'): a freeform text, added to the annotation\n - 'id' ('str'): the instance id of the note\n - 'type' ('str'): the type of note\n - 'target' ('str'): the id of the related annotation\n - 'note' ('str'): the text body of the note", "### Usage\n\nThe 'brat' dataset script can be used by calling 'load_dataset()' method and passing any arguments that are accepted by the 'BratConfig' (which is a special BuilderConfig). It requires at least the 'url' argument. The full list of arguments is as follows:\n\n- 'url' ('str'): the url of the dataset which should point to either a zip file or a directory containing the Brat data ('*.txt') and annotation ('*.ann') files\n\n- 'description' ('str', optional): the description of the dataset\n\n- 'citation' ('str', optional): the citation of the dataset\n\n- 'homepage' ('str', optional): the homepage of the dataset\n\n- 'split_paths' ('dict', optional): a mapping of (arbitrary) split names to subdirectories or lists of files (without extension), e.g. '{\"train\": \"path/to/train_directory\", \"test\": \"path/to/test_director\"}' or '{\"train\": [\"path/to/train_file1\", \"path/to/train_file2\"]}'. In both cases (subdirectory paths or file paths), the paths are relative to the url. If 'split_paths' is not provided, the dataset will be loaded from the root directory and all direct subfolders will be considered as splits.\n\n- 'file_name_blacklist' ('list', optional): a list of file names (without extension) that should be ignored, e.g. '[\"A28\"]'. This is useful if the dataset contains files that are not valid brat files.\n\nImportant: Using the 'data_dir' parameter of the 'load_dataset()' method overrides the 'url' parameter of the 'BratConfig'.\n\nWe provide an example of SciArg dataset below:", "## Additional Information", "### Licensing Information\n\n\\\\]" ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for [kejian/pile-severetoxic-balanced2]", "## Generation Procedures \n\nThe dataset was constructed using documents from the Pile scored using Perspective API SEVERE-TOXICITY scores.\n\nThe procedure was the following:\n- The first half of this dataset is kejian/pile-severetoxic-chunk-0, 100k most toxic documents from Pile chunk-0\n- The second half of this dataset is kejian/pile-severetoxic-random100k, 100k randomly sampled documents from Pile chunk-3\n- Then, the dataset was shuffled and a 9:1 train-test split was done", "## Basic Statistics \n\nThe average scores of the most toxic and random half are 0.555 and 0.061, respectively. The average score of the whole dataset is 0.308; the median is 0.385.\n\n is 0.337. The correlation between score and document length is 0.099" ]

[ "TAGS\n#region-us \n", "# Dataset Card for [kejian/pile-severetoxic-balanced2]", "## Generation Procedures \n\nThe dataset was constructed using documents from the Pile scored using Perspective API SEVERE-TOXICITY scores.\n\nThe procedure was the following:\n- The first half of this dataset is kejian/pile-severetoxic-chunk-0, 100k most toxic documents from Pile chunk-0\n- The second half of this dataset is kejian/pile-severetoxic-random100k, 100k randomly sampled documents from Pile chunk-3\n- Then, the dataset was shuffled and a 9:1 train-test split was done", "## Basic Statistics \n\nThe average scores of the most toxic and random half are 0.555 and 0.061, respectively. The average score of the whole dataset is 0.308; the median is 0.385.\n\n is 0.337. The correlation between score and document length is 0.099" ]

[ "arxiv:1910.10261", "arxiv:2106.10161", "region:us" ]

[ "1910.10261", "2106.10161" ]

[]

[ "### Audio files in opus format", "### Audio files in wav format\n\n\nManifest files with all the training transcription texts are in the train\\_crowd9.tar archive listed in the table:\n\n\nArchives: train\\_farfield.tar, Size: 15.4 GB, Links: URL\nArchives: train\\_crowd0.tar, Size: 11 GB, Links: URL\nArchives: train\\_crowd1.tar, Size: 14 GB, Links: URL\nArchives: train\\_crowd2.tar, Size: 13.2 GB, Links: URL\nArchives: train\\_crowd3.tar, Size: 11.6 GB, Links: URL\nArchives: train\\_crowd4.tar, Size: 15.8 GB, Links: URL\nArchives: train\\_crowd5.tar, Size: 13.1 GB, Links: URL\nArchives: train\\_crowd6.tar, Size: 15.7 GB, Links: URL\nArchives: train\\_crowd7.tar, Size: 12.7 GB, Links: URL\nArchives: train\\_crowd8.tar, Size: 12.2 GB, Links: URL\nArchives: train\\_crowd9.tar, Size: 8.08 GB, Links: URL\nArchives: URL, Size: 1.3 GB, Links: URL", "### Acoustic and language models\n\n\nAcoustic model built using QuartzNet15x5 architecture and trained using NeMo toolkit\n\n\nThree n-gram language models created using KenLM Language Model Toolkit\n\n\n* LM built on Common Crawl Russian dataset\n* LM built on Golos train set\n* LM built on Common Crawl and Golos datasets together (50/50)\n\n\nArchives: QuartzNet15x5\\_golos.nemo, Size: 68 MB, Links: URL\nArchives: URL, Size: 4.8 GB, Links: URL\n\n\nGolos data and models are also available in the hub of pre-trained models, datasets, and containers - DataHub ML Space. You can train the model and deploy it on the high-performance SberCloud infrastructure in ML Space - full-cycle machine learning development platform for DS-teams collaboration based on the Christofari Supercomputer.\n\n\nEvaluation\n----------\n\n\nPercents of Word Error Rate for different test sets\n\n\n\n1 Common Voice - Mozilla's initiative to help teach machines how real people speak.\n\n\nResources\n---------\n\n\n[[URL] Golos: Russian Dataset for Speech Research](URL\n\n\n[[URL] Golos — самый большой русскоязычный речевой датасет, размеченный вручную, теперь в открытом доступе](URL\n\n\n[[URL] Как улучшить распознавание русской речи до 3% WER с помощью открытых данных](URL" ]

[ "TAGS\n#arxiv-1910.10261 #arxiv-2106.10161 #region-us \n", "### Audio files in opus format", "### Audio files in wav format\n\n\nManifest files with all the training transcription texts are in the train\\_crowd9.tar archive listed in the table:\n\n\nArchives: train\\_farfield.tar, Size: 15.4 GB, Links: URL\nArchives: train\\_crowd0.tar, Size: 11 GB, Links: URL\nArchives: train\\_crowd1.tar, Size: 14 GB, Links: URL\nArchives: train\\_crowd2.tar, Size: 13.2 GB, Links: URL\nArchives: train\\_crowd3.tar, Size: 11.6 GB, Links: URL\nArchives: train\\_crowd4.tar, Size: 15.8 GB, Links: URL\nArchives: train\\_crowd5.tar, Size: 13.1 GB, Links: URL\nArchives: train\\_crowd6.tar, Size: 15.7 GB, Links: URL\nArchives: train\\_crowd7.tar, Size: 12.7 GB, Links: URL\nArchives: train\\_crowd8.tar, Size: 12.2 GB, Links: URL\nArchives: train\\_crowd9.tar, Size: 8.08 GB, Links: URL\nArchives: URL, Size: 1.3 GB, Links: URL", "### Acoustic and language models\n\n\nAcoustic model built using QuartzNet15x5 architecture and trained using NeMo toolkit\n\n\nThree n-gram language models created using KenLM Language Model Toolkit\n\n\n* LM built on Common Crawl Russian dataset\n* LM built on Golos train set\n* LM built on Common Crawl and Golos datasets together (50/50)\n\n\nArchives: QuartzNet15x5\\_golos.nemo, Size: 68 MB, Links: URL\nArchives: URL, Size: 4.8 GB, Links: URL\n\n\nGolos data and models are also available in the hub of pre-trained models, datasets, and containers - DataHub ML Space. You can train the model and deploy it on the high-performance SberCloud infrastructure in ML Space - full-cycle machine learning development platform for DS-teams collaboration based on the Christofari Supercomputer.\n\n\nEvaluation\n----------\n\n\nPercents of Word Error Rate for different test sets\n\n\n\n1 Common Voice - Mozilla's initiative to help teach machines how real people speak.\n\n\nResources\n---------\n\n\n[[URL] Golos: Russian Dataset for Speech Research](URL\n\n\n[[URL] Golos — самый большой русскоязычный речевой датасет, размеченный вручную, теперь в открытом доступе](URL\n\n\n[[URL] Как улучшить распознавание русской речи до 3% WER с помощью открытых данных](URL" ]

[ "task_categories:token-classification", "task_ids:named-entity-recognition", "task_ids:part-of-speech", "annotations_creators:hired_annotators", "language_creators:found", "multilinguality:monolingual", "size_categories:10<n<10K", "language:pl", "license:other", "region:us" ]

[]

[ "pl" ]

[]

[ "TAGS\n#task_categories-token-classification #task_ids-named-entity-recognition #task_ids-part-of-speech #annotations_creators-hired_annotators #language_creators-found #multilinguality-monolingual #size_categories-10<n<10K #language-Polish #license-other #region-us \n" ]

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:multilingual", "size_categories:10K<n<100K", "source_datasets:original", "arxiv:2006.07235", "arxiv:2004.02192", "arxiv:1908.04531", "arxiv:2004.14454", "arxiv:2003.07459", "region:us" ]

[ "2006.07235", "2004.02192", "1908.04531", "2004.14454", "2003.07459" ]

[]

[ "### Dataset Summary\n\n\nOffensEval 2020 features a multilingual dataset with five languages. The languages included in OffensEval 2020 are:\n\n\n* Arabic\n* Danish\n* English\n* Greek\n* Turkish\n\n\nThe annotation follows the hierarchical tagset proposed in the Offensive Language Identification Dataset (OLID) and used in OffensEval 2019.\nIn this taxonomy we break down offensive content into the following three sub-tasks taking the type and target of offensive content into account.\nThe following sub-tasks were organized:\n\n\n* Sub-task A - Offensive language identification;\n* Sub-task B - Automatic categorization of offense types;\n* Sub-task C - Offense target identification.\n\n\nEnglish training data is omitted so needs to be collected otherwise (see URL\n\n\nThe source datasets come from:\n\n\n* Arabic URL URL\n* Danish URL URL/URL\n* English URL URL\n* Greek URL URL\n* Turkish URL", "### Supported Tasks and Leaderboards\n\n\n* OffensEval 2020", "### Languages\n\n\nFive are covered: bcp47 'ar;da;en;gr;tr'\n\n\nDataset Structure\n-----------------\n\n\nThere are five named configs, one per language:\n\n\n* 'ar' Arabic\n* 'da' Danish\n* 'en' English\n* 'gr' Greek\n* 'tr' Turkish\n\n\nThe training data for English is absent - this is 9M tweets that need to be rehydrated on their own. See URL", "### Data Instances\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature.\n* 'text': a 'string'.\n* 'subtask\\_a': whether or not the instance is offensive; '0: NOT, 1: OFF'", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nCollecting data for abusive language classification. Different rational for each dataset.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nVaries per language dataset", "#### Who are the source language producers?\n\n\nSocial media users", "### Annotations", "#### Annotation process\n\n\nVaries per language dataset", "#### Who are the annotators?\n\n\nVaries per language dataset; native speakers", "### Personal and Sensitive Information\n\n\nThe data was public at the time of collection. No PII removal has been performed.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe data definitely contains abusive language. The data could be used to develop and propagate offensive language against every target group involved, i.e. ableism, racism, sexism, ageism, and so on.", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe datasets is curated by each sub-part's paper authors.", "### Licensing Information\n\n\nThis data is available and distributed under Creative Commons attribution license, CC-BY 4.0.", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-expert-generated #language_creators-found #multilinguality-multilingual #size_categories-10K<n<100K #source_datasets-original #arxiv-2006.07235 #arxiv-2004.02192 #arxiv-1908.04531 #arxiv-2004.14454 #arxiv-2003.07459 #region-us \n", "### Dataset Summary\n\n\nOffensEval 2020 features a multilingual dataset with five languages. The languages included in OffensEval 2020 are:\n\n\n* Arabic\n* Danish\n* English\n* Greek\n* Turkish\n\n\nThe annotation follows the hierarchical tagset proposed in the Offensive Language Identification Dataset (OLID) and used in OffensEval 2019.\nIn this taxonomy we break down offensive content into the following three sub-tasks taking the type and target of offensive content into account.\nThe following sub-tasks were organized:\n\n\n* Sub-task A - Offensive language identification;\n* Sub-task B - Automatic categorization of offense types;\n* Sub-task C - Offense target identification.\n\n\nEnglish training data is omitted so needs to be collected otherwise (see URL\n\n\nThe source datasets come from:\n\n\n* Arabic URL URL\n* Danish URL URL/URL\n* English URL URL\n* Greek URL URL\n* Turkish URL", "### Supported Tasks and Leaderboards\n\n\n* OffensEval 2020", "### Languages\n\n\nFive are covered: bcp47 'ar;da;en;gr;tr'\n\n\nDataset Structure\n-----------------\n\n\nThere are five named configs, one per language:\n\n\n* 'ar' Arabic\n* 'da' Danish\n* 'en' English\n* 'gr' Greek\n* 'tr' Turkish\n\n\nThe training data for English is absent - this is 9M tweets that need to be rehydrated on their own. See URL", "### Data Instances\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature.\n* 'text': a 'string'.\n* 'subtask\\_a': whether or not the instance is offensive; '0: NOT, 1: OFF'", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nCollecting data for abusive language classification. Different rational for each dataset.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nVaries per language dataset", "#### Who are the source language producers?\n\n\nSocial media users", "### Annotations", "#### Annotation process\n\n\nVaries per language dataset", "#### Who are the annotators?\n\n\nVaries per language dataset; native speakers", "### Personal and Sensitive Information\n\n\nThe data was public at the time of collection. No PII removal has been performed.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe data definitely contains abusive language. The data could be used to develop and propagate offensive language against every target group involved, i.e. ableism, racism, sexism, ageism, and so on.", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe datasets is curated by each sub-part's paper authors.", "### Licensing Information\n\n\nThis data is available and distributed under Creative Commons attribution license, CC-BY 4.0.", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:no-annotation", "language_creators:expert-generated", "multilinguality:multilingual", "size_categories:n<1K", "source_datasets:original", "license:mit", "region:us" ]

[]

[]

[ "# Dataset Card for HONEST", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper: URL\n- Point of Contact: Debora Nozza", "### Dataset Summary\n\nHONEST dataset comprises a set of templates for measuring hurtful sentence completions in language models. The templates are provided in six languages (English, Italian, French, Portuguese, Romanian, and Spanish) for binary gender and in English for LGBTQAI+ individuals.\nWARNING: This dataset contains content that are offensive and/or hateful in nature.", "### Languages\nEnglish, Italian, French, Portuguese, Romanian, and Spanish.", "## Dataset Structure", "### Data Instances\nAn example of an English instance:", "### Data Fields\n\n- 'template_masked': the template\n- 'category': category of the template useful for aggregations\n- 'identity': identity term used to fill the templates\n- 'number': singular or plural version of the identity term\n- 'raw': the raw template\n- 'type': the template type (occupation, descriptive_adjective, or descriptive_verb)", "### Data Splits\n\nThere is no data splits. HONEST dataset should not be used as training but just as a test dataset.", "## Dataset Creation", "### Curation Rationale\n\nLarge language models (LLMs) have revolutionized the field of NLP. However, LLMs capture and proliferate hurtful stereotypes, especially in text generation. HONEST permits to measure hurtful sentence completion of language models in different languages and for different targets.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe manually generate a set of these templates for all the languages. Note that we also cover gender-inflected languages.", "#### Who are the source language producers?\n\nTemplates were generated by native speakers of the respective languages from European Countries, all in the age group 25-30.", "### Personal and Sensitive Information\n\nThe data we share is not sensitive to personal information, as it does not contain information about individuals.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThe dataset permits to quantify the amount of hurtful completions in language models. Researchers and practitioners can use this contribution to understand if a model is safe to use or not.", "### Discussion of Biases\n\nThe choice of the templates is arbitrary.", "### Other Known Limitations\n\nWe want to explicitly address the limitation of our approach with respect to the binary nature of our gender analysis for the languages other than English.", "## Additional Information", "### Dataset Curators\n\n- Debora Nozza - URL@URL\n- Federico Bianchi - f.bianchi@URL\n- Dirk Hovy - URL@URL", "### Licensing Information\n\nMIT License", "### Contributions\nThanks to @dnozza for adding this dataset." ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-no-annotation #language_creators-expert-generated #multilinguality-multilingual #size_categories-n<1K #source_datasets-original #license-mit #region-us \n", "# Dataset Card for HONEST", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper: URL\n- Point of Contact: Debora Nozza", "### Dataset Summary\n\nHONEST dataset comprises a set of templates for measuring hurtful sentence completions in language models. The templates are provided in six languages (English, Italian, French, Portuguese, Romanian, and Spanish) for binary gender and in English for LGBTQAI+ individuals.\nWARNING: This dataset contains content that are offensive and/or hateful in nature.", "### Languages\nEnglish, Italian, French, Portuguese, Romanian, and Spanish.", "## Dataset Structure", "### Data Instances\nAn example of an English instance:", "### Data Fields\n\n- 'template_masked': the template\n- 'category': category of the template useful for aggregations\n- 'identity': identity term used to fill the templates\n- 'number': singular or plural version of the identity term\n- 'raw': the raw template\n- 'type': the template type (occupation, descriptive_adjective, or descriptive_verb)", "### Data Splits\n\nThere is no data splits. HONEST dataset should not be used as training but just as a test dataset.", "## Dataset Creation", "### Curation Rationale\n\nLarge language models (LLMs) have revolutionized the field of NLP. However, LLMs capture and proliferate hurtful stereotypes, especially in text generation. HONEST permits to measure hurtful sentence completion of language models in different languages and for different targets.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe manually generate a set of these templates for all the languages. Note that we also cover gender-inflected languages.", "#### Who are the source language producers?\n\nTemplates were generated by native speakers of the respective languages from European Countries, all in the age group 25-30.", "### Personal and Sensitive Information\n\nThe data we share is not sensitive to personal information, as it does not contain information about individuals.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThe dataset permits to quantify the amount of hurtful completions in language models. Researchers and practitioners can use this contribution to understand if a model is safe to use or not.", "### Discussion of Biases\n\nThe choice of the templates is arbitrary.", "### Other Known Limitations\n\nWe want to explicitly address the limitation of our approach with respect to the binary nature of our gender analysis for the languages other than English.", "## Additional Information", "### Dataset Curators\n\n- Debora Nozza - URL@URL\n- Federico Bianchi - f.bianchi@URL\n- Dirk Hovy - URL@URL", "### Licensing Information\n\nMIT License", "### Contributions\nThanks to @dnozza for adding this dataset." ]

[ "task_categories:automatic-speech-recognition", "multilinguality:multilingual", "language:en", "language:de", "language:fr", "language:es", "language:pl", "language:it", "language:ro", "language:hu", "language:cs", "language:nl", "language:fi", "language:hr", "language:sk", "language:sl", "language:et", "language:lt", "license:cc0-1.0", "license:other", "arxiv:2101.00390", "region:us" ]

[ "2101.00390" ]

[ "en", "de", "fr", "es", "pl", "it", "ro", "hu", "cs", "nl", "fi", "hr", "sk", "sl", "et", "lt" ]

[ "### Dataset Summary\n\n\nVoxPopuli is a large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation.\nThe raw data is collected from 2009-2020 European Parliament event recordings. We acknowledge the European Parliament for creating and sharing these materials.\nThis implementation contains transcribed speech data for 18 languages.\nIt also contains 29 hours of transcribed speech data of non-native English intended for research in ASR for accented speech (15 L2 accents)", "### Example usage\n\n\nVoxPopuli contains labelled data for 18 languages. To load a specific language pass its name as a config name:\n\n\nTo load all the languages in a single dataset use \"multilang\" config name:\n\n\nTo load a specific set of languages, use \"multilang\" config name and pass a list of required languages to 'languages' parameter:\n\n\nTo load accented English data, use \"en\\_accented\" config name:\n\n\nNote that L2 English subset contains only 'test' split.", "### Supported Tasks and Leaderboards\n\n\n* automatic-speech-recognition: The dataset can be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER).\n\n\nAccented English subset can also be used for research in ASR for accented speech (15 L2 accents)", "### Languages\n\n\nVoxPopuli contains labelled (transcribed) data for 18 languages:\n\n\n\nAccented speech transcribed data has 15 various L2 accents:\n\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'audio\\_id' (string) - id of audio segment\n* 'language' (datasets.ClassLabel) - numerical id of audio segment\n* 'audio' (datasets.Audio) - a dictionary containing the path to the audio, the decoded audio array, and the sampling rate. In non-streaming mode (default), the path points to the locally extracted audio. In streaming mode, the path is the relative path of an audio inside its archive (as files are not downloaded and extracted locally).\n* 'raw\\_text' (string) - original (orthographic) audio segment text\n* 'normalized\\_text' (string) - normalized audio segment transcription\n* 'gender' (string) - gender of speaker\n* 'speaker\\_id' (string) - id of speaker\n* 'is\\_gold\\_transcript' (bool) - ?\n* 'accent' (string) - type of accent, for example \"en\\_lt\", if applicable, else \"None\".", "### Data Splits\n\n\nAll configs (languages) except for accented English contain data in three splits: train, validation and test. Accented English 'en\\_accented' config contains only test split.\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data\n\n\nThe raw data is collected from 2009-2020 European Parliament event recordings", "#### Initial Data Collection and Normalization\n\n\nThe VoxPopuli transcribed set comes from aligning the full-event source speech audio with the transcripts for plenary sessions. Official timestamps\nare available for locating speeches by speaker in the full session, but they are frequently inaccurate, resulting in truncation of the speech or mixture\nof fragments from the preceding or the succeeding speeches. To calibrate the original timestamps,\nwe perform speaker diarization (SD) on the full-session audio using URL (Bredin et al.2020) and adopt the nearest SD timestamps (by L1 distance to the original ones) instead for segmentation.\nFull-session audios are segmented into speech paragraphs by speaker, each of which has a transcript available.\n\n\nThe speech paragraphs have an average duration of 197 seconds, which leads to significant. We hence further segment these paragraphs into utterances with a\nmaximum duration of 20 seconds. We leverage speech recognition (ASR) systems to force-align speech paragraphs to the given transcripts.\nThe ASR systems are TDS models (Hannun et al., 2019) trained with ASG criterion (Collobert et al., 2016) on audio tracks from in-house deidentified video data.\n\n\nThe resulting utterance segments may have incorrect transcriptions due to incomplete raw transcripts or inaccurate ASR force-alignment.\nWe use the predictions from the same ASR systems as references and filter the candidate segments by a maximum threshold of 20% character error rate(CER).", "#### Who are the source language producers?\n\n\nSpeakers are participants of the European Parliament events, many of them are EU officials.", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases\n\n\nGender speakers distribution is imbalanced, percentage of female speakers is mostly lower than 50% across languages, with the minimum of 15% for the Lithuanian language data.\n\n\nVoxPopuli includes all available speeches from the 2009-2020 EP events without any selections on the topics or speakers.\nThe speech contents represent the standpoints of the speakers in the EP events, many of which are EU officials.", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThe dataset is distributet under CC0 license, see also European Parliament's legal notice for the raw data.\n\n\nPlease cite this paper:", "### Contributions\n\n\nThanks to @polinaeterna for adding this dataset." ]

[ "TAGS\n#task_categories-automatic-speech-recognition #multilinguality-multilingual #language-English #language-German #language-French #language-Spanish #language-Polish #language-Italian #language-Romanian #language-Hungarian #language-Czech #language-Dutch #language-Finnish #language-Croatian #language-Slovak #language-Slovenian #language-Estonian #language-Lithuanian #license-cc0-1.0 #license-other #arxiv-2101.00390 #region-us \n", "### Dataset Summary\n\n\nVoxPopuli is a large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation.\nThe raw data is collected from 2009-2020 European Parliament event recordings. We acknowledge the European Parliament for creating and sharing these materials.\nThis implementation contains transcribed speech data for 18 languages.\nIt also contains 29 hours of transcribed speech data of non-native English intended for research in ASR for accented speech (15 L2 accents)", "### Example usage\n\n\nVoxPopuli contains labelled data for 18 languages. To load a specific language pass its name as a config name:\n\n\nTo load all the languages in a single dataset use \"multilang\" config name:\n\n\nTo load a specific set of languages, use \"multilang\" config name and pass a list of required languages to 'languages' parameter:\n\n\nTo load accented English data, use \"en\\_accented\" config name:\n\n\nNote that L2 English subset contains only 'test' split.", "### Supported Tasks and Leaderboards\n\n\n* automatic-speech-recognition: The dataset can be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER).\n\n\nAccented English subset can also be used for research in ASR for accented speech (15 L2 accents)", "### Languages\n\n\nVoxPopuli contains labelled (transcribed) data for 18 languages:\n\n\n\nAccented speech transcribed data has 15 various L2 accents:\n\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'audio\\_id' (string) - id of audio segment\n* 'language' (datasets.ClassLabel) - numerical id of audio segment\n* 'audio' (datasets.Audio) - a dictionary containing the path to the audio, the decoded audio array, and the sampling rate. In non-streaming mode (default), the path points to the locally extracted audio. In streaming mode, the path is the relative path of an audio inside its archive (as files are not downloaded and extracted locally).\n* 'raw\\_text' (string) - original (orthographic) audio segment text\n* 'normalized\\_text' (string) - normalized audio segment transcription\n* 'gender' (string) - gender of speaker\n* 'speaker\\_id' (string) - id of speaker\n* 'is\\_gold\\_transcript' (bool) - ?\n* 'accent' (string) - type of accent, for example \"en\\_lt\", if applicable, else \"None\".", "### Data Splits\n\n\nAll configs (languages) except for accented English contain data in three splits: train, validation and test. Accented English 'en\\_accented' config contains only test split.\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data\n\n\nThe raw data is collected from 2009-2020 European Parliament event recordings", "#### Initial Data Collection and Normalization\n\n\nThe VoxPopuli transcribed set comes from aligning the full-event source speech audio with the transcripts for plenary sessions. Official timestamps\nare available for locating speeches by speaker in the full session, but they are frequently inaccurate, resulting in truncation of the speech or mixture\nof fragments from the preceding or the succeeding speeches. To calibrate the original timestamps,\nwe perform speaker diarization (SD) on the full-session audio using URL (Bredin et al.2020) and adopt the nearest SD timestamps (by L1 distance to the original ones) instead for segmentation.\nFull-session audios are segmented into speech paragraphs by speaker, each of which has a transcript available.\n\n\nThe speech paragraphs have an average duration of 197 seconds, which leads to significant. We hence further segment these paragraphs into utterances with a\nmaximum duration of 20 seconds. We leverage speech recognition (ASR) systems to force-align speech paragraphs to the given transcripts.\nThe ASR systems are TDS models (Hannun et al., 2019) trained with ASG criterion (Collobert et al., 2016) on audio tracks from in-house deidentified video data.\n\n\nThe resulting utterance segments may have incorrect transcriptions due to incomplete raw transcripts or inaccurate ASR force-alignment.\nWe use the predictions from the same ASR systems as references and filter the candidate segments by a maximum threshold of 20% character error rate(CER).", "#### Who are the source language producers?\n\n\nSpeakers are participants of the European Parliament events, many of them are EU officials.", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases\n\n\nGender speakers distribution is imbalanced, percentage of female speakers is mostly lower than 50% across languages, with the minimum of 15% for the Lithuanian language data.\n\n\nVoxPopuli includes all available speeches from the 2009-2020 EP events without any selections on the topics or speakers.\nThe speech contents represent the standpoints of the speakers in the EP events, many of which are EU officials.", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThe dataset is distributet under CC0 license, see also European Parliament's legal notice for the raw data.\n\n\nPlease cite this paper:", "### Contributions\n\n\nThanks to @polinaeterna for adding this dataset." ]

[ "license:odc-by", "region:us" ]

[]

[]

[ "# MuP - Multi Perspective Scientific Document Summarization\n\nGenerating summaries of scientific documents is known to be a challenging task. Majority of existing work in summarization assumes only one single best gold summary for each given document. Having only one gold summary negatively impacts our ability to evaluate the quality of summarization systems as writing summaries is a subjective activity. At the same time, annotating multiple gold summaries for scientific documents can be extremely expensive as it requires domain experts to read and understand long scientific documents. This shared task will enable exploring methods for generating multi-perspective summaries. We introduce a novel summarization corpus, leveraging data from scientific peer reviews to capture diverse perspectives from the reader's point of view." ]

[ "TAGS\n#license-odc-by #region-us \n", "# MuP - Multi Perspective Scientific Document Summarization\n\nGenerating summaries of scientific documents is known to be a challenging task. Majority of existing work in summarization assumes only one single best gold summary for each given document. Having only one gold summary negatively impacts our ability to evaluate the quality of summarization systems as writing summaries is a subjective activity. At the same time, annotating multiple gold summaries for scientific documents can be extremely expensive as it requires domain experts to read and understand long scientific documents. This shared task will enable exploring methods for generating multi-perspective summaries. We introduce a novel summarization corpus, leveraging data from scientific peer reviews to capture diverse perspectives from the reader's point of view." ]

[ "task_categories:question-answering", "task_categories:visual-question-answering", "task_ids:closed-domain-qa", "annotations_creators:expert-generated", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:other", "arxiv:1604.02748", "region:us" ]

[ "1604.02748" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThe Tumblr GIF (TGIF) dataset contains 100K animated GIFs and 120K sentences describing visual content of the animated GIFs. The animated GIFs have been collected from Tumblr, from randomly selected posts published between May and June of 2015. We provide the URLs of animated GIFs in this release. The sentences are collected via crowdsourcing, with a carefully designed annotation interface that ensures high quality dataset. We provide one sentence per animated GIF for the training and validation splits, and three sentences per GIF for the test split. The dataset shall be used to evaluate animated GIF/video description techniques.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_path': 'str' \"URL\n-'video\\_bytes': 'large\\_bytes' video file in bytes format\n* 'en\\_global\\_captions': 'list\\_str' List of english captions describing the entire video", "### Data Splits", "### Annotations\n\n\nQuoting TGIF paper: \n\n\"We annotated animated GIFs with natural language descriptions using the crowdsourcing service CrowdFlower.\nWe carefully designed our annotation task with various\nquality control mechanisms to ensure the sentences are both\nsyntactically and semantically of high quality.\nA total of 931 workers participated in our annotation\ntask. We allowed workers only from Australia, Canada, New Zealand, UK and USA in an effort to collect fluent descriptions from native English speakers. Figure 2 shows the\ninstructions given to the workers. Each task showed 5 animated GIFs and asked the worker to describe each with one\nsentence. To promote language style diversity, each worker\ncould rate no more than 800 images (0.7% of our corpus).\nWe paid 0.02 USD per sentence; the entire crowdsourcing\ncost less than 4K USD. We provide details of our annotation\ntask in the supplementary material.\"", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThis dataset is provided to be used for approved non-commercial research purposes. No personally identifying information is available in this dataset.", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "TAGS\n#task_categories-question-answering #task_categories-visual-question-answering #task_ids-closed-domain-qa #annotations_creators-expert-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-other #arxiv-1604.02748 #region-us \n", "### Dataset Summary\n\n\nThe Tumblr GIF (TGIF) dataset contains 100K animated GIFs and 120K sentences describing visual content of the animated GIFs. The animated GIFs have been collected from Tumblr, from randomly selected posts published between May and June of 2015. We provide the URLs of animated GIFs in this release. The sentences are collected via crowdsourcing, with a carefully designed annotation interface that ensures high quality dataset. We provide one sentence per animated GIF for the training and validation splits, and three sentences per GIF for the test split. The dataset shall be used to evaluate animated GIF/video description techniques.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_path': 'str' \"URL\n-'video\\_bytes': 'large\\_bytes' video file in bytes format\n* 'en\\_global\\_captions': 'list\\_str' List of english captions describing the entire video", "### Data Splits", "### Annotations\n\n\nQuoting TGIF paper: \n\n\"We annotated animated GIFs with natural language descriptions using the crowdsourcing service CrowdFlower.\nWe carefully designed our annotation task with various\nquality control mechanisms to ensure the sentences are both\nsyntactically and semantically of high quality.\nA total of 931 workers participated in our annotation\ntask. We allowed workers only from Australia, Canada, New Zealand, UK and USA in an effort to collect fluent descriptions from native English speakers. Figure 2 shows the\ninstructions given to the workers. Each task showed 5 animated GIFs and asked the worker to describe each with one\nsentence. To promote language style diversity, each worker\ncould rate no more than 800 images (0.7% of our corpus).\nWe paid 0.02 USD per sentence; the entire crowdsourcing\ncost less than 4K USD. We provide details of our annotation\ntask in the supplementary material.\"", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThis dataset is provided to be used for approved non-commercial research purposes. No personally identifying information is available in this dataset.", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "task_categories:text-classification", "annotations_creators:found", "language_creators:found", "multilinguality:multilingual", "size_categories:100K<n<1M", "source_datasets:original", "language:da", "language:nn", "language:nb", "language:fo", "language:is", "language:sv", "license:cc-by-sa-3.0", "language-identification", "region:us" ]

[]

[ "da", "nn", "nb", "fo", "is", "sv" ]

[ "# Dataset Card for nordic_langid", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: URL\n- Leaderboard: \n- Point of Contact: René Haas", "### Dataset Summary\n\nAutomatic language identification is a challenging problem. Discriminating\nbetween closely related languages is especially difficult. This paper presents\na machine learning approach for automatic language identification for the\nNordic languages, which often suffer miscategorisation by existing \nstate-of-the-art tools. Concretely we will focus on discrimination between six \nNordic language: Danish, Swedish, Norwegian (Nynorsk), Norwegian (Bokmål), \nFaroese and Icelandic.\n\nThis is the data for the tasks. Two variants are provided: 10K and 50K, with\nholding 10,000 and 50,000 examples for each language respectively.\n\nFor more info, see the paper: Discriminating Between Similar Nordic Languages.", "### Supported Tasks and Leaderboards\n\n*", "### Languages\n\nThis dataset is in six similar Nordic language:\n\n- Danish, 'da'\n- Faroese, 'fo'\n- Icelandic, 'is'\n- Norwegian Bokmål, 'nb'\n- Norwegian Nynorsk, 'nn'\n- Swedish, 'sv'", "## Dataset Structure\n\nThe dataset has two parts, one with 10K samples per language and another with 50K per language.\nThe original splits and data allocation used in the paper is presented here.", "### Data Instances", "### Data Fields\n\n- 'id': the sentence's unique identifier, 'string'\n- 'sentence': the test to be classifier, a 'string'\n- 'language': the class, one of 'da', 'fo', 'is', 'nb', 'no', 'sv'.", "### Data Splits\n\nTrain and Test splits are provided, divided using the code provided with the paper.", "## Dataset Creation", "### Curation Rationale\n\nData is taken from Wikipedia and Tatoeba from each of these six languages.", "### Source Data", "#### Initial Data Collection and Normalization\n\nData collection Data was scraped from Wikipedia. We downloaded summaries for randomly chosen Wikipedia\narticles in each of the languages, saved as raw text\nto six .txt files of about 10MB each.\nThe 50K section is extended with Tatoeba data, which provides a different register to Wikipedia text, and then topped up with more Wikipedia data.\n\nExtracting Sentences The first pass in sentence\ntokenisation is splitting by line breaks. We then extract shorter sentences with the sentence tokenizer\n(sent_tokenize) function from NLTK (Loper\nand Bird, 2002). This does a better job than just\nsplitting by ’.’ due to the fact that abbreviations,\nwhich can appear in a legitimate sentence, typically\ninclude a period symbol.\n\nCleaning characters The initial data set has\nmany characters that do not belong to the alphabets of the languages we work with. Often the\nWikipedia pages for people or places contain names\nin foreign languages. For example a summary\nmight contain Chinese or Russian characters which\nare not strong signals for the purpose of discriminating between the target languages.\nFurther, it can be that some characters in the\ntarget languages are mis-encoded. These misencodings are also not likely to be intrinsically\nstrong or stable signals.\nTo simplify feature extraction, and to reduce the\nsize of the vocabulary, the raw data is converted\nto lowercase and stripped of all characters which\nare not part of the standard alphabet of the six\nlanguages using a character whitelist.", "#### Who are the source language producers?\n\nThe source language is from Wikipedia contributors and Tatoeba contributors.", "### Annotations", "#### Annotation process\n\nThe annotations were found.", "#### Who are the annotators?\n\nThe annotations were found. They are determined by which language section a contributor posts their content to.", "### Personal and Sensitive Information\n\nThe data hasn't been checked for PII, and is already all public. Tatoeba is is based on translations of synthetic conversational turns and is unlikely to bear personal or sensitive information.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended to help correctly identify content in the languages of six minority languages. Existing systems often confuse these, especially Bokmål and Danish or Icelandic and Faroese. However, some dialects are missed (for example Bornholmsk) and the closed nature of the classification task thus excludes speakers of these languages without recognising their existence.", "### Discussion of Biases\n\nThe text comes from only two genres, so might not transfer well to other domains.", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\nThe data here is licensed CC-BY-SA 3.0. If you use this data, you MUST state its origin." ]

[ "TAGS\n#task_categories-text-classification #annotations_creators-found #language_creators-found #multilinguality-multilingual #size_categories-100K<n<1M #source_datasets-original #language-Danish #language-Norwegian Nynorsk #language-Norwegian Bokmål #language-Faroese #language-Icelandic #language-Swedish #license-cc-by-sa-3.0 #language-identification #region-us \n", "# Dataset Card for nordic_langid", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: URL\n- Leaderboard: \n- Point of Contact: René Haas", "### Dataset Summary\n\nAutomatic language identification is a challenging problem. Discriminating\nbetween closely related languages is especially difficult. This paper presents\na machine learning approach for automatic language identification for the\nNordic languages, which often suffer miscategorisation by existing \nstate-of-the-art tools. Concretely we will focus on discrimination between six \nNordic language: Danish, Swedish, Norwegian (Nynorsk), Norwegian (Bokmål), \nFaroese and Icelandic.\n\nThis is the data for the tasks. Two variants are provided: 10K and 50K, with\nholding 10,000 and 50,000 examples for each language respectively.\n\nFor more info, see the paper: Discriminating Between Similar Nordic Languages.", "### Supported Tasks and Leaderboards\n\n*", "### Languages\n\nThis dataset is in six similar Nordic language:\n\n- Danish, 'da'\n- Faroese, 'fo'\n- Icelandic, 'is'\n- Norwegian Bokmål, 'nb'\n- Norwegian Nynorsk, 'nn'\n- Swedish, 'sv'", "## Dataset Structure\n\nThe dataset has two parts, one with 10K samples per language and another with 50K per language.\nThe original splits and data allocation used in the paper is presented here.", "### Data Instances", "### Data Fields\n\n- 'id': the sentence's unique identifier, 'string'\n- 'sentence': the test to be classifier, a 'string'\n- 'language': the class, one of 'da', 'fo', 'is', 'nb', 'no', 'sv'.", "### Data Splits\n\nTrain and Test splits are provided, divided using the code provided with the paper.", "## Dataset Creation", "### Curation Rationale\n\nData is taken from Wikipedia and Tatoeba from each of these six languages.", "### Source Data", "#### Initial Data Collection and Normalization\n\nData collection Data was scraped from Wikipedia. We downloaded summaries for randomly chosen Wikipedia\narticles in each of the languages, saved as raw text\nto six .txt files of about 10MB each.\nThe 50K section is extended with Tatoeba data, which provides a different register to Wikipedia text, and then topped up with more Wikipedia data.\n\nExtracting Sentences The first pass in sentence\ntokenisation is splitting by line breaks. We then extract shorter sentences with the sentence tokenizer\n(sent_tokenize) function from NLTK (Loper\nand Bird, 2002). This does a better job than just\nsplitting by ’.’ due to the fact that abbreviations,\nwhich can appear in a legitimate sentence, typically\ninclude a period symbol.\n\nCleaning characters The initial data set has\nmany characters that do not belong to the alphabets of the languages we work with. Often the\nWikipedia pages for people or places contain names\nin foreign languages. For example a summary\nmight contain Chinese or Russian characters which\nare not strong signals for the purpose of discriminating between the target languages.\nFurther, it can be that some characters in the\ntarget languages are mis-encoded. These misencodings are also not likely to be intrinsically\nstrong or stable signals.\nTo simplify feature extraction, and to reduce the\nsize of the vocabulary, the raw data is converted\nto lowercase and stripped of all characters which\nare not part of the standard alphabet of the six\nlanguages using a character whitelist.", "#### Who are the source language producers?\n\nThe source language is from Wikipedia contributors and Tatoeba contributors.", "### Annotations", "#### Annotation process\n\nThe annotations were found.", "#### Who are the annotators?\n\nThe annotations were found. They are determined by which language section a contributor posts their content to.", "### Personal and Sensitive Information\n\nThe data hasn't been checked for PII, and is already all public. Tatoeba is is based on translations of synthetic conversational turns and is unlikely to bear personal or sensitive information.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended to help correctly identify content in the languages of six minority languages. Existing systems often confuse these, especially Bokmål and Danish or Icelandic and Faroese. However, some dialects are missed (for example Bornholmsk) and the closed nature of the classification task thus excludes speakers of these languages without recognising their existence.", "### Discussion of Biases\n\nThe text comes from only two genres, so might not transfer well to other domains.", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\nThe data here is licensed CC-BY-SA 3.0. If you use this data, you MUST state its origin." ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for HowTo100M", "## Table of Contents\nTable of Contents\nDataset Description\n Dataset Summary\n Dataset Preprocessing\n Supported Tasks and Leaderboards\n Languages\nDataset Structure\n Data Instances\n Data Fields\n Data Splits\nDataset Creation\n Curation Rationale\n Source Data\n Annotations\n Personal and Sensitive Information\nConsiderations for Using the Data\n Social Impact of Dataset\n Discussion of Biases\n Other Known Limitations\nAdditional Information\n Dataset Curators\n Licensing Information\n Citation Information\n Contributions", "## Dataset Description\n\nHomepage: HowTo100M homepage\nRepository: Github repo\nPaper: HowTo100M: Learning a Text-Video Embedding by Watching Hundred Million Narrated Video Clips\nPoint of Contact: Antoine Miech", "### Dataset Summary\n\nHowTo100M is a large-scale dataset of narrated videos with an emphasis on instructional videos where content creators teach complex tasks with an explicit intention of explaining the visual content on screen. HowTo100M features a total of:\n136M video clips with captions sourced from 1.2M Youtube videos (15 years of video)\n23k activities from domains such as cooking, hand crafting, personal care, gardening or fitness\n\nEach video is associated with a narration available as subtitles automatically downloaded from Youtube.", "### Dataset Preprocessing\n\nThis dataset does not contain the videos by default. You would need to follow the instructions here from the dataset creators and fill out a form to get a userd id and a password to download the videos from their server.\n\nOnce you have these two, you can fetch the videos by mapping the following function to the 'path' column:", "### Supported Tasks and Leaderboards\n\n'video-to-text': This dataset can be used to train a model for Video Captioning where the goal is to predict a caption given the video.", "### Languages\n\nAll captions are in English and are either coming from available YouTube subtitles (manually written) or the output of an Automatic Speech Recognition system.", "## Dataset Structure", "### Data Instances\n\nEach instance in HowTo100M represents a single video with two lists of start and end of segments and a caption for each segment.", "### Data Fields\n\n'video_id': YouTube video ID\n'path': Path to download the videos from the authors once proper access is accredited\n'category_1': Highest level task category from WikiHow\n'category_2': Second highest level task category from WikiHow\n'rank': YouTube serach result rank of the video when querying the task\n'starts': List corresponding to the end timestamps of each segment\n'ends': List corresponding to the end timestamps of each segment\n'captions': List of all the captions (one per segment)", "### Data Splits\n\nAll the data is contained in training split. The training set has 1M instances.", "## Dataset Creation", "### Curation Rationale\n\nFrom the paper:\n> we first start by acquiring a large list of activities using WikiHow1 – an online resource that contains 120,000 articles on How to ... for a variety of domains ranging from cooking to human relationships structured in a hierarchy. We are primarily interested in “visual tasks” that involve some interaction with the physical world (e.g. Making peanut butter, Pruning a tree) as compared to others that are more abstract (e.g. Ending a toxic relationship, Choosing a gift). To obtain predominantly visual tasks, we limit them to one of 12 categories (listed in Table 2). We exclude categories such as Relationships and Finance and Business, that may be more abstract. We further refine the set of tasks, by filtering them in a semi-automatic way. In particular, we restrict the primary verb to physical actions, such as make, build and change, and discard non-physical verbs, such as be, accept and feel. This procedure yields 23,611 visual tasks in total.\n\n> We search for YouTube videos related to the task by forming a query with how to preceding the task name (e.g. how to paint furniture). We choose videos that have English subtitles either uploaded manually, generated automatically by YouTube ASR, or generated automatically after translation from a different language by YouTube API. We improve the quality and consistency of the dataset, by adopting the following criteria. We restrict to the top 200 search results, as the latter ones may not be related to the query task. Videos with less than 100 views are removed as they are often of poor quality or are amateurish. We also ignore videos that have less than 100 words as that may be insufficient text to learn a good video-language embedding. Finally, we remove videos longer than 2,000 seconds. As some videos may appear in several tasks, we deduplicate videos based on YouTube IDs. However, note that the dataset may still contain duplicates if a video was uploaded several times or edited and re-uploaded. Nevertheless, this is not a concern at our scale.", "### Source Data\n\nThe source videos come from YouTube.", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?\n\nYouTube uploaders.", "### Annotations", "#### Annotation process\n\nSubtitles are generated or manually written. Note that the narrated captions have been processed. In fact, authors have removed a significant number of stop words\nwhich are not relevant for the learning of the text-video joint embedding. The list of stop words can be found here: URL You can find the unprocessed caption file (i.e. with stop words) here.", "#### Who are the annotators?\n\nYouTube uploaders or machine-generated outputs.", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nAntoine Miech, Dimitri Zhukov, Jean-Baptiste Alayrac, Makarand Tapaswi, Ivan Laptev, Josef Sivic", "### Licensing Information\n\nNot specified.", "### Contributions\n\nThanks to @VictorSanh for adding this dataset." ]

[ "TAGS\n#region-us \n", "# Dataset Card for HowTo100M", "## Table of Contents\nTable of Contents\nDataset Description\n Dataset Summary\n Dataset Preprocessing\n Supported Tasks and Leaderboards\n Languages\nDataset Structure\n Data Instances\n Data Fields\n Data Splits\nDataset Creation\n Curation Rationale\n Source Data\n Annotations\n Personal and Sensitive Information\nConsiderations for Using the Data\n Social Impact of Dataset\n Discussion of Biases\n Other Known Limitations\nAdditional Information\n Dataset Curators\n Licensing Information\n Citation Information\n Contributions", "## Dataset Description\n\nHomepage: HowTo100M homepage\nRepository: Github repo\nPaper: HowTo100M: Learning a Text-Video Embedding by Watching Hundred Million Narrated Video Clips\nPoint of Contact: Antoine Miech", "### Dataset Summary\n\nHowTo100M is a large-scale dataset of narrated videos with an emphasis on instructional videos where content creators teach complex tasks with an explicit intention of explaining the visual content on screen. HowTo100M features a total of:\n136M video clips with captions sourced from 1.2M Youtube videos (15 years of video)\n23k activities from domains such as cooking, hand crafting, personal care, gardening or fitness\n\nEach video is associated with a narration available as subtitles automatically downloaded from Youtube.", "### Dataset Preprocessing\n\nThis dataset does not contain the videos by default. You would need to follow the instructions here from the dataset creators and fill out a form to get a userd id and a password to download the videos from their server.\n\nOnce you have these two, you can fetch the videos by mapping the following function to the 'path' column:", "### Supported Tasks and Leaderboards\n\n'video-to-text': This dataset can be used to train a model for Video Captioning where the goal is to predict a caption given the video.", "### Languages\n\nAll captions are in English and are either coming from available YouTube subtitles (manually written) or the output of an Automatic Speech Recognition system.", "## Dataset Structure", "### Data Instances\n\nEach instance in HowTo100M represents a single video with two lists of start and end of segments and a caption for each segment.", "### Data Fields\n\n'video_id': YouTube video ID\n'path': Path to download the videos from the authors once proper access is accredited\n'category_1': Highest level task category from WikiHow\n'category_2': Second highest level task category from WikiHow\n'rank': YouTube serach result rank of the video when querying the task\n'starts': List corresponding to the end timestamps of each segment\n'ends': List corresponding to the end timestamps of each segment\n'captions': List of all the captions (one per segment)", "### Data Splits\n\nAll the data is contained in training split. The training set has 1M instances.", "## Dataset Creation", "### Curation Rationale\n\nFrom the paper:\n> we first start by acquiring a large list of activities using WikiHow1 – an online resource that contains 120,000 articles on How to ... for a variety of domains ranging from cooking to human relationships structured in a hierarchy. We are primarily interested in “visual tasks” that involve some interaction with the physical world (e.g. Making peanut butter, Pruning a tree) as compared to others that are more abstract (e.g. Ending a toxic relationship, Choosing a gift). To obtain predominantly visual tasks, we limit them to one of 12 categories (listed in Table 2). We exclude categories such as Relationships and Finance and Business, that may be more abstract. We further refine the set of tasks, by filtering them in a semi-automatic way. In particular, we restrict the primary verb to physical actions, such as make, build and change, and discard non-physical verbs, such as be, accept and feel. This procedure yields 23,611 visual tasks in total.\n\n> We search for YouTube videos related to the task by forming a query with how to preceding the task name (e.g. how to paint furniture). We choose videos that have English subtitles either uploaded manually, generated automatically by YouTube ASR, or generated automatically after translation from a different language by YouTube API. We improve the quality and consistency of the dataset, by adopting the following criteria. We restrict to the top 200 search results, as the latter ones may not be related to the query task. Videos with less than 100 views are removed as they are often of poor quality or are amateurish. We also ignore videos that have less than 100 words as that may be insufficient text to learn a good video-language embedding. Finally, we remove videos longer than 2,000 seconds. As some videos may appear in several tasks, we deduplicate videos based on YouTube IDs. However, note that the dataset may still contain duplicates if a video was uploaded several times or edited and re-uploaded. Nevertheless, this is not a concern at our scale.", "### Source Data\n\nThe source videos come from YouTube.", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?\n\nYouTube uploaders.", "### Annotations", "#### Annotation process\n\nSubtitles are generated or manually written. Note that the narrated captions have been processed. In fact, authors have removed a significant number of stop words\nwhich are not relevant for the learning of the text-video joint embedding. The list of stop words can be found here: URL You can find the unprocessed caption file (i.e. with stop words) here.", "#### Who are the annotators?\n\nYouTube uploaders or machine-generated outputs.", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nAntoine Miech, Dimitri Zhukov, Jean-Baptiste Alayrac, Makarand Tapaswi, Ivan Laptev, Josef Sivic", "### Licensing Information\n\nNot specified.", "### Contributions\n\nThanks to @VictorSanh for adding this dataset." ]

[ "region:us" ]

[]

[]

[ "### Languages\n\n\nThe BCP-47 code for the dataset's language is unk.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nA sample from this dataset looks as follows:", "### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):", "### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ "TAGS\n#region-us \n", "### Languages\n\n\nThe BCP-47 code for the dataset's language is unk.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nA sample from this dataset looks as follows:", "### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):", "### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ "task_categories:other", "annotations_creators:crowdsourced", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:1K<n<10K", "source_datasets:original", "language:en", "license:other", "arxiv:1604.01753", "region:us" ]

[ "1604.01753" ]

[ "en" ]

[ "### Dataset Summary\n\n\nCharades is dataset composed of 9848 videos of daily indoors activities collected through Amazon Mechanical Turk. 267 different users were presented with a sentence, that includes objects and actions from a fixed vocabulary, and they recorded a video acting out the sentence (like in a game of Charades). The dataset contains 66,500 temporal annotations for 157 action classes, 41,104 labels for 46 object classes, and 27,847 textual descriptions of the videos", "### Supported Tasks and Leaderboards\n\n\n* 'multilabel-action-classification': The goal of this task is to classify actions happening in a video. This is a multilabel classification. The leaderboard is available here", "### Languages\n\n\nThe annotations in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'video\\_id': 'str' Unique identifier for each video.\n* 'video': 'str' Path to the video file\n* 'subject': 'str' Unique identifier for each subject in the dataset\n* 'scene': 'str' One of 15 indoor scenes in the dataset, such as Kitchen\n* 'quality': 'int' The quality of the video judged by an annotator (7-point scale, 7=high quality), -100 if missing\n* 'relevance': 'int' The relevance of the video to the script judged by an annotated (7-point scale, 7=very relevant), -100 if missing\n* 'verified': 'str' 'Yes' if an annotator successfully verified that the video matches the script, else 'No'\n* 'script': 'str' The human-generated script used to generate the video\n* 'descriptions': 'List[str]' List of descriptions by annotators watching the video\n* 'labels': 'List[int]' Multi-label actions found in the video. Indices from 0 to 156.\n* 'action\\_timings': 'List[Tuple[int, int]]' Timing where each of the above actions happened.\n* 'length': 'float' The length of the video in seconds\n\n\n\n\n Click here to see the full list of Charades class labels mapping:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\n\n> \n> Computer vision has a great potential to help our daily lives by searching for lost keys, watering flowers or reminding us to take a pill. To succeed with such tasks, computer vision methods need to be trained from real and diverse examples of our daily dynamic scenes. While most of such scenes are not particularly exciting, they typically do not appear on YouTube, in movies or TV broadcasts. So how do we collect sufficiently many diverse but boring samples representing our lives? We propose a novel Hollywood in Homes approach to collect such data. Instead of shooting videos in the lab, we ensure diversity by distributing and crowdsourcing the whole process of video creation from script writing to video recording and annotation.\n> \n> \n>", "### Source Data", "#### Initial Data Collection and Normalization\n\n\n\n> \n> Similar to filming, we have a three-step process for generating a video. The first step is generating the script of the indoor video. The key here is to allow workers to generate diverse scripts yet ensure that we have enough data for each category. The second step in the process is to use the script and ask workers to record a video of that sentence being acted out. In the final step, we ask the workers to verify if the recorded video corresponds to script, followed by an annotation procedure.\n> \n> \n>", "#### Who are the source language producers?\n\n\nAmazon Mechnical Turk annotators", "### Annotations", "#### Annotation process\n\n\n\n> \n> Similar to filming, we have a three-step process for generating a video. The first step is generating the script of the indoor video. The key here is to allow workers to generate diverse scripts yet ensure that we have enough data for each category. The second step in the process is to use the script and ask workers to record a video of that sentence being acted out. In the final step, we ask the workers to verify if the recorded video corresponds to script, followed by an annotation procedure.\n> \n> \n>", "#### Who are the annotators?\n\n\nAmazon Mechnical Turk annotators", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nAMT annotators", "### Licensing Information\n\n\nLicense for Non-Commercial Use\n\n\nIf this software is redistributed, this license must be included. The term software includes any source files, documentation, executables, models, and data.\n\n\nThis software and data is available for general use by academic or non-profit, or government-sponsored researchers. It may also be used for evaluation purposes elsewhere. This license does not grant the right to use this software or any derivation of it in a for-profit enterprise. For commercial use, please contact The Allen Institute for Artificial Intelligence.\n\n\nThis license does not grant the right to modify and publicly release the data in any form.\n\n\nThis license does not grant the right to distribute the data to a third party in any form.\n\n\nThe subjects in this data should be treated with respect and dignity. This license only grants the right to publish short segments or still images in an academic publication where necessary to present examples, experimental results, or observations.\n\n\nThis software comes with no warranty or guarantee of any kind. By using this software, the user accepts full liability.\n\n\nThe Allen Institute for Artificial Intelligence (C) 2016.", "### Contributions\n\n\nThanks to @apsdehal for adding this dataset." ]

[ "TAGS\n#task_categories-other #annotations_creators-crowdsourced #language_creators-crowdsourced #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-English #license-other #arxiv-1604.01753 #region-us \n", "### Dataset Summary\n\n\nCharades is dataset composed of 9848 videos of daily indoors activities collected through Amazon Mechanical Turk. 267 different users were presented with a sentence, that includes objects and actions from a fixed vocabulary, and they recorded a video acting out the sentence (like in a game of Charades). The dataset contains 66,500 temporal annotations for 157 action classes, 41,104 labels for 46 object classes, and 27,847 textual descriptions of the videos", "### Supported Tasks and Leaderboards\n\n\n* 'multilabel-action-classification': The goal of this task is to classify actions happening in a video. This is a multilabel classification. The leaderboard is available here", "### Languages\n\n\nThe annotations in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'video\\_id': 'str' Unique identifier for each video.\n* 'video': 'str' Path to the video file\n* 'subject': 'str' Unique identifier for each subject in the dataset\n* 'scene': 'str' One of 15 indoor scenes in the dataset, such as Kitchen\n* 'quality': 'int' The quality of the video judged by an annotator (7-point scale, 7=high quality), -100 if missing\n* 'relevance': 'int' The relevance of the video to the script judged by an annotated (7-point scale, 7=very relevant), -100 if missing\n* 'verified': 'str' 'Yes' if an annotator successfully verified that the video matches the script, else 'No'\n* 'script': 'str' The human-generated script used to generate the video\n* 'descriptions': 'List[str]' List of descriptions by annotators watching the video\n* 'labels': 'List[int]' Multi-label actions found in the video. Indices from 0 to 156.\n* 'action\\_timings': 'List[Tuple[int, int]]' Timing where each of the above actions happened.\n* 'length': 'float' The length of the video in seconds\n\n\n\n\n Click here to see the full list of Charades class labels mapping:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\n\n> \n> Computer vision has a great potential to help our daily lives by searching for lost keys, watering flowers or reminding us to take a pill. To succeed with such tasks, computer vision methods need to be trained from real and diverse examples of our daily dynamic scenes. While most of such scenes are not particularly exciting, they typically do not appear on YouTube, in movies or TV broadcasts. So how do we collect sufficiently many diverse but boring samples representing our lives? We propose a novel Hollywood in Homes approach to collect such data. Instead of shooting videos in the lab, we ensure diversity by distributing and crowdsourcing the whole process of video creation from script writing to video recording and annotation.\n> \n> \n>", "### Source Data", "#### Initial Data Collection and Normalization\n\n\n\n> \n> Similar to filming, we have a three-step process for generating a video. The first step is generating the script of the indoor video. The key here is to allow workers to generate diverse scripts yet ensure that we have enough data for each category. The second step in the process is to use the script and ask workers to record a video of that sentence being acted out. In the final step, we ask the workers to verify if the recorded video corresponds to script, followed by an annotation procedure.\n> \n> \n>", "#### Who are the source language producers?\n\n\nAmazon Mechnical Turk annotators", "### Annotations", "#### Annotation process\n\n\n\n> \n> Similar to filming, we have a three-step process for generating a video. The first step is generating the script of the indoor video. The key here is to allow workers to generate diverse scripts yet ensure that we have enough data for each category. The second step in the process is to use the script and ask workers to record a video of that sentence being acted out. In the final step, we ask the workers to verify if the recorded video corresponds to script, followed by an annotation procedure.\n> \n> \n>", "#### Who are the annotators?\n\n\nAmazon Mechnical Turk annotators", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nAMT annotators", "### Licensing Information\n\n\nLicense for Non-Commercial Use\n\n\nIf this software is redistributed, this license must be included. The term software includes any source files, documentation, executables, models, and data.\n\n\nThis software and data is available for general use by academic or non-profit, or government-sponsored researchers. It may also be used for evaluation purposes elsewhere. This license does not grant the right to use this software or any derivation of it in a for-profit enterprise. For commercial use, please contact The Allen Institute for Artificial Intelligence.\n\n\nThis license does not grant the right to modify and publicly release the data in any form.\n\n\nThis license does not grant the right to distribute the data to a third party in any form.\n\n\nThe subjects in this data should be treated with respect and dignity. This license only grants the right to publish short segments or still images in an academic publication where necessary to present examples, experimental results, or observations.\n\n\nThis software comes with no warranty or guarantee of any kind. By using this software, the user accepts full liability.\n\n\nThe Allen Institute for Artificial Intelligence (C) 2016.", "### Contributions\n\n\nThanks to @apsdehal for adding this dataset." ]

[ "license:apache-2.0", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-apache-2.0 #region-us \n" ]

[ "region:us" ]

[]

[]

[ "# Italian Tweets Test Dataset\nThis is a test dataset that is available for debugging reasons only. It contains errors. Please do not use.", "## How to Use" ]

[ "TAGS\n#region-us \n", "# Italian Tweets Test Dataset\nThis is a test dataset that is available for debugging reasons only. It contains errors. Please do not use.", "## How to Use" ]

[ "task_categories:translation", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:translation", "size_categories:1K<n<10K", "source_datasets:original", "license:cc-by-4.0", "region:us" ]

[]

[ "da", "da-bornholm" ]

[ "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: URL\n- Point of Contact: Leon Derczynski\n- Size of downloaded dataset files: 490 KB\n- Size of the generated dataset: 582 KB\n- Total amount of disk used: 1072 KB", "### Dataset Summary\n\nThis dataset is parallel text for Bornholmsk and Danish. \n\nFor more details, see the paper Bornholmsk Natural Language Processing: Resources and Tools.", "### Supported Tasks and Leaderboards\n\n*", "### Languages\n\nBornholmsk, a language variant of Danish spoken on the island of Bornholm, and Danish. bcp47: 'da-bornholm' and 'da-DK'", "## Dataset Structure", "### Data Instances", "### Data Fields\n\n'id': the sentence ID, 'int'\n'da-bornholm': the Bornholmsk text, 'string'\n'da': the Danish translation, 'string'", "### Data Splits\n\n* Train: 5785 sentence pairs\n* Validation: 500 sentence pairs\n* Test: 500 sentence pairs", "## Dataset Creation", "### Curation Rationale\n\nTo gather as much parallel Bornholmsk together as possible", "### Source Data", "#### Initial Data Collection and Normalization\n\nFrom a translation of Kuhre's Sansager, a selection of colloquial resources, and a prototype Bornholmsk/Danish dictionary", "#### Who are the source language producers?\n\nNative speakers of Bornholmsk who have produced works in their native language, or translated them to Danish. Much of the data is the result of a community of Bornholmsk speakers volunteering their time across the island in an effort to capture this endangered language.", "### Annotations", "#### Annotation process\n\nNo annotations", "#### Who are the annotators?\n\nNative speakers of Bornholmsk, mostly aged 60+.", "### Personal and Sensitive Information\n\nUnknown, but low risk of presence, given the source material", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThe hope behind this data is to enable people to learn and use Bornholmsk", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis collection of Bornholmsk is curated by Leon Derczynski and Alex Speed Kjeldsen", "### Licensing Information\n\nCreative Commons Attribution 4.0" ]

[ "TAGS\n#task_categories-translation #annotations_creators-expert-generated #language_creators-found #multilinguality-translation #size_categories-1K<n<10K #source_datasets-original #license-cc-by-4.0 #region-us \n", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: URL\n- Point of Contact: Leon Derczynski\n- Size of downloaded dataset files: 490 KB\n- Size of the generated dataset: 582 KB\n- Total amount of disk used: 1072 KB", "### Dataset Summary\n\nThis dataset is parallel text for Bornholmsk and Danish. \n\nFor more details, see the paper Bornholmsk Natural Language Processing: Resources and Tools.", "### Supported Tasks and Leaderboards\n\n*", "### Languages\n\nBornholmsk, a language variant of Danish spoken on the island of Bornholm, and Danish. bcp47: 'da-bornholm' and 'da-DK'", "## Dataset Structure", "### Data Instances", "### Data Fields\n\n'id': the sentence ID, 'int'\n'da-bornholm': the Bornholmsk text, 'string'\n'da': the Danish translation, 'string'", "### Data Splits\n\n* Train: 5785 sentence pairs\n* Validation: 500 sentence pairs\n* Test: 500 sentence pairs", "## Dataset Creation", "### Curation Rationale\n\nTo gather as much parallel Bornholmsk together as possible", "### Source Data", "#### Initial Data Collection and Normalization\n\nFrom a translation of Kuhre's Sansager, a selection of colloquial resources, and a prototype Bornholmsk/Danish dictionary", "#### Who are the source language producers?\n\nNative speakers of Bornholmsk who have produced works in their native language, or translated them to Danish. Much of the data is the result of a community of Bornholmsk speakers volunteering their time across the island in an effort to capture this endangered language.", "### Annotations", "#### Annotation process\n\nNo annotations", "#### Who are the annotators?\n\nNative speakers of Bornholmsk, mostly aged 60+.", "### Personal and Sensitive Information\n\nUnknown, but low risk of presence, given the source material", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThe hope behind this data is to enable people to learn and use Bornholmsk", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis collection of Bornholmsk is curated by Leon Derczynski and Alex Speed Kjeldsen", "### Licensing Information\n\nCreative Commons Attribution 4.0" ]

[ "language:en", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#language-English #region-us \n" ]

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "language:da", "license:other", "not-for-all-audiences", "region:us" ]

[]

[ "da" ]

[ "### Dataset Summary\n\n\nThis is a high-quality dataset of annotated posts sampled from social\nmedia posts and annotated for misogyny. Danish language.\n\n\nOnline misogyny, a category of online abusive language, has serious and\nharmful social consequences. Automatic detection of misogynistic language\nonline, while imperative, poses complicated challenges to both data\ngathering, data annotation, and bias mitigation, as this type of data is\nlinguistically complex and diverse.\n\n\nSee the accompanying ACL paper Annotating Online Misogyny for full details.", "### Supported Tasks and Leaderboards\n\n\n*", "### Languages\n\n\nDanish ('bcp47:da')\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### Bajer\n\n\n* Size of downloaded dataset files: 7.29 MiB\n* Size of the generated dataset: 6.57 MiB\n* Total amount of disk used: 13.85 MiB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature, unique identifier in this dataset.\n* 'dataset\\_id': a 'string' feature, internal annotation identifier.\n* 'label\\_id': a 'string' feature, internal annotation sequence number.\n* 'text': a 'string' of the text that's annotated.\n* 'sampling': a 'string' describing which sampling technique surfaced this message\n* 'subtask\\_A': is the text abusive 'ABUS' or not 'NOT'? '0: NOT, 1: ABUS'\n* 'subtask\\_B': for abusive text, what's the target - individual 'IND', group 'GRP', other 'OTH', or untargeted 'UNT'? '0: IND, 1: GRP, 2: OTH, 3: UNT, 4: not applicable'\n* 'subtask\\_C1': for group-targeted abuse, what's the group - misogynistic 'SEX', other 'OTH', or racist 'RAC'? '0: SEX, 1: OTH, 2: RAC, 3: not applicable'\n* 'subtask\\_C2': for misogyny, is it neosexist 'NEOSEX', discrediting 'DISCREDIT', normative stereotyping 'NOR', benevolent sexism 'AMBIVALENT', dominance 'DOMINANCE', or harassment 'HARASSMENT'? '0: NEOSEX, 1: DISCREDIT, 2: NOR, 3: AMBIVALENT, 4: DOMINANCE, 5: HARASSMENT, 6: not applicable'", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nThe goal was to collect data for developing an annotation schema of online misogyny.\n\n\nRandom sampling of text often results in scarcity of examples of specifically misogynistic content (e.g. (Wulczyn et al., 2017;\nFounta et al., 2018)). Therefore, we used the common alternative of collecting data by using predefined keywords with a potentially high search hit\n(e.g. Waseem and Hovy (2016)), and identifying\nrelevant user-profiles (e.g. (Anzovino et al., 2018))\nand related topics (e.g. (Kumar et al., 2018)).\n\n\nWe searched for keywords (specific slurs, hashtags), that are known to occur in sexist posts. These\nwere defined by previous work, a slur list from\nReddit, and from interviews and surveys of online\nmisogyny among women. We also searched for\nbroader terms like “sex” or “women”, which do\nnot appear exclusively in a misogynistic context,\nfor example in the topic search, where we gathered\nrelevant posts and their comments from the social\nmedia pages of public media. A complete list of\nkeywords can be found in the appendix.\n\n\nSocial media provides a potentially biased, but\nbroad snapshot of online human discourse, with\nplenty of language and behaviours represented. Following best practice guidelines (Vidgen and Derczynski, 2020), we sampled from a language for\nwhich there are no existing annotations of the target\nphenomenon: Danish.\n\n\nDifferent social media platforms attract different user groups and can exhibit domain-specific\nlanguage (Karan and Snajder ˇ , 2018). Rather than\nchoosing one platform (existing misogyny datasets\nare primarily based on Twitter and Reddit (Guest\net al., 2021)), we sampled from multiple platforms:\nStatista (2020) shows that the platform where most\nDanish users are present is Facebook, followed\nby Twitter, YouTube, Instagram and lastly, Reddit.\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text. Data was gathered based on: keyword-based search (i.e. purposive sampling); topic-based search; and content from specific users.", "#### Who are the source language producers?\n\n\nDanish-speaking social media users", "### Annotations", "#### Annotation process\n\n\nIn annotating our dataset, we built on the MATTER\nframework (Pustejovsky and Stubbs, 2012) and use\nthe variation presented by Finlayson and Erjavec\n(2017) (the MALER framework), where the Train & Test stages are replaced by Leveraging of annotations for one’s particular goal, in our case the\ncreation of a comprehensive taxonomy.\n\n\nWe created a set of guidelines for the annotators.\nThe annotators were first asked to read the guidelines and individually annotate about 150 different\nposts, after which there was a shared discussion.\nAfter this pilot round, the volume of samples per annotator was increased and every sample labeled by\n2-3 annotators. When instances were ‘flagged’ or\nannotators disagreed on them, they were discussed\nduring weekly meetings, and misunderstandings\nwere resolved together with the external facilitator. After round three, when reaching 7k annotated\nposts (Figure 2), we continued with independent\nannotations maintaining a 15% instance overlap\nbetween randomly picked annotator pairs.\n\n\nManagement of annotator disagreement is an important part of the process design. Disagreements\ncan be solved by majority voting (Davidson et al.,\n2017; Wiegand et al., 2019), labeled as abuse if at\nleast one annotator has labeled it (Golbeck et al.,\n2017) or by a third objective instance (Gao and\nHuang, 2017). Most datasets use crowdsourcing\nplatforms or a few academic experts for annotation\n(Vidgen and Derczynski, 2020). Inter-annotatoragreement (IAA) and classification performance\nare established as two grounded evaluation measurements for annotation quality (Vidgen and Derczynski, 2020). Comparing the performance of amateur annotators (while providing guidelines) with\nexpert annotators for sexism and racism annotation,\nWaseem (2016) show that the quality of amateur\nannotators is competitive with expert annotations\nwhen several amateurs agree. Facing the trade-off\nbetween training annotators intensely and the number of involved annotators, we continued with the\ntrained annotators and group discussions/ individual revisions for flagged content and disagreements\n(Section 5.4).", "#### Who are the annotators?\n\n\n---|---\nGender|6 female, 2 male (8 total)\nAge:| 5 <30; 3 ≥30\nEthnicity:| 5 Danish: 1 Persian, 1 Arabic, 1 Polish\nStudy/occupation: | Linguistics (2); Health/Software Design; Ethnography/Digital Design; Communication/Psychology; Anthropology/Broadcast Moderator; Ethnography/Climate Change; Film Artist", "### Personal and Sensitive Information\n\n\nUsernames and PII were stripped during annotation process by skipping content containing these and eliding it from the final dataset\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe data contains abusive language. It may be possible to identify original speakers based on the content, so the data is only available for research purposes under a restrictive license and conditions. We hope that identifying sexism can help moderators. There is a possibility that the content here could be used to generate misogyny in Danish, which would place women in Denmark in an even more hostile environment, and for this reason data access is restricted and tracked.", "### Discussion of Biases\n\n\nWe have taken pains to mitigate as many biases as we were aware of in this work.\n\n\nSelection biases: Selection biases for abusive\nlanguage can be seen in the sampling of text, for instance when using keyword search (Wiegand et al.,\n2019), topic dependency (Ousidhoum et al., 2020), users (Wiegand et al., 2019), domain (Wiegand\net al., 2019), time (Florio et al., 2020) and lack of\nlinguistic variety (Vidgen and Derczynski, 2020).\n\n\nLabel biases: Label biases can be caused by, for\ninstance, non-representative annotator selection,\nlack in training/domain expertise, preconceived\nnotions, or pre-held stereotypes. These biases are\ntreated in relation to abusive language datasets\nby several sources, e.g. general sampling and\nannotators biases (Waseem, 2016; Al Kuwatly\net al., 2020), biases towards minority identity\nmentions based for example on gender or race\n(Davidson et al., 2017; Dixon et al., 2018; Park\net al., 2018; Davidson et al., 2019), and political\nannotator biases (Wich et al., 2020). Other qualitative biases comprise, for instance, demographic\nbias, over-generalization, topic exposure as social\nbiases (Hovy and Spruit, 2016).\n\n\nWe applied several measures to mitigate biases\noccurring through the annotation design and execution: First, we selected labels grounded in existing,\npeer-reviewed research from more than one field.\nSecond, we aimed for diversity in annotator profiles\nin terms of age, gender, dialect, and background.\nThird, we recruited a facilitator with a background\nin ethnographic studies and provided intense annotator training. Fourth, we engaged in weekly group\ndiscussions, iteratively improving the codebook\nand integrating edge cases. Fifth, the selection of\nplatforms from which we sampled data is based on\nlocal user representation in Denmark, rather than\nconvenience. Sixth, diverse sampling methods for\ndata collection reduced selection biases.", "### Other Known Limitations\n\n\nThe data is absolutely NOT a reasonable or in any way stratified sample of social media text, so class prevalence/balance here says nothing about incidences of these phenomena in the wild. That said, we hypothesis that the distribution of types of misogyny in this data (subtask C2) is roughly representative of how misogyny presents on the studied platforms.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors and the ethnographer-led annotation team.", "### Licensing Information\n\n\nThe data is licensed under a restrictive usage agreement. Apply for access here", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-Danish #license-other #not-for-all-audiences #region-us \n", "### Dataset Summary\n\n\nThis is a high-quality dataset of annotated posts sampled from social\nmedia posts and annotated for misogyny. Danish language.\n\n\nOnline misogyny, a category of online abusive language, has serious and\nharmful social consequences. Automatic detection of misogynistic language\nonline, while imperative, poses complicated challenges to both data\ngathering, data annotation, and bias mitigation, as this type of data is\nlinguistically complex and diverse.\n\n\nSee the accompanying ACL paper Annotating Online Misogyny for full details.", "### Supported Tasks and Leaderboards\n\n\n*", "### Languages\n\n\nDanish ('bcp47:da')\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### Bajer\n\n\n* Size of downloaded dataset files: 7.29 MiB\n* Size of the generated dataset: 6.57 MiB\n* Total amount of disk used: 13.85 MiB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature, unique identifier in this dataset.\n* 'dataset\\_id': a 'string' feature, internal annotation identifier.\n* 'label\\_id': a 'string' feature, internal annotation sequence number.\n* 'text': a 'string' of the text that's annotated.\n* 'sampling': a 'string' describing which sampling technique surfaced this message\n* 'subtask\\_A': is the text abusive 'ABUS' or not 'NOT'? '0: NOT, 1: ABUS'\n* 'subtask\\_B': for abusive text, what's the target - individual 'IND', group 'GRP', other 'OTH', or untargeted 'UNT'? '0: IND, 1: GRP, 2: OTH, 3: UNT, 4: not applicable'\n* 'subtask\\_C1': for group-targeted abuse, what's the group - misogynistic 'SEX', other 'OTH', or racist 'RAC'? '0: SEX, 1: OTH, 2: RAC, 3: not applicable'\n* 'subtask\\_C2': for misogyny, is it neosexist 'NEOSEX', discrediting 'DISCREDIT', normative stereotyping 'NOR', benevolent sexism 'AMBIVALENT', dominance 'DOMINANCE', or harassment 'HARASSMENT'? '0: NEOSEX, 1: DISCREDIT, 2: NOR, 3: AMBIVALENT, 4: DOMINANCE, 5: HARASSMENT, 6: not applicable'", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nThe goal was to collect data for developing an annotation schema of online misogyny.\n\n\nRandom sampling of text often results in scarcity of examples of specifically misogynistic content (e.g. (Wulczyn et al., 2017;\nFounta et al., 2018)). Therefore, we used the common alternative of collecting data by using predefined keywords with a potentially high search hit\n(e.g. Waseem and Hovy (2016)), and identifying\nrelevant user-profiles (e.g. (Anzovino et al., 2018))\nand related topics (e.g. (Kumar et al., 2018)).\n\n\nWe searched for keywords (specific slurs, hashtags), that are known to occur in sexist posts. These\nwere defined by previous work, a slur list from\nReddit, and from interviews and surveys of online\nmisogyny among women. We also searched for\nbroader terms like “sex” or “women”, which do\nnot appear exclusively in a misogynistic context,\nfor example in the topic search, where we gathered\nrelevant posts and their comments from the social\nmedia pages of public media. A complete list of\nkeywords can be found in the appendix.\n\n\nSocial media provides a potentially biased, but\nbroad snapshot of online human discourse, with\nplenty of language and behaviours represented. Following best practice guidelines (Vidgen and Derczynski, 2020), we sampled from a language for\nwhich there are no existing annotations of the target\nphenomenon: Danish.\n\n\nDifferent social media platforms attract different user groups and can exhibit domain-specific\nlanguage (Karan and Snajder ˇ , 2018). Rather than\nchoosing one platform (existing misogyny datasets\nare primarily based on Twitter and Reddit (Guest\net al., 2021)), we sampled from multiple platforms:\nStatista (2020) shows that the platform where most\nDanish users are present is Facebook, followed\nby Twitter, YouTube, Instagram and lastly, Reddit.\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text. Data was gathered based on: keyword-based search (i.e. purposive sampling); topic-based search; and content from specific users.", "#### Who are the source language producers?\n\n\nDanish-speaking social media users", "### Annotations", "#### Annotation process\n\n\nIn annotating our dataset, we built on the MATTER\nframework (Pustejovsky and Stubbs, 2012) and use\nthe variation presented by Finlayson and Erjavec\n(2017) (the MALER framework), where the Train & Test stages are replaced by Leveraging of annotations for one’s particular goal, in our case the\ncreation of a comprehensive taxonomy.\n\n\nWe created a set of guidelines for the annotators.\nThe annotators were first asked to read the guidelines and individually annotate about 150 different\nposts, after which there was a shared discussion.\nAfter this pilot round, the volume of samples per annotator was increased and every sample labeled by\n2-3 annotators. When instances were ‘flagged’ or\nannotators disagreed on them, they were discussed\nduring weekly meetings, and misunderstandings\nwere resolved together with the external facilitator. After round three, when reaching 7k annotated\nposts (Figure 2), we continued with independent\nannotations maintaining a 15% instance overlap\nbetween randomly picked annotator pairs.\n\n\nManagement of annotator disagreement is an important part of the process design. Disagreements\ncan be solved by majority voting (Davidson et al.,\n2017; Wiegand et al., 2019), labeled as abuse if at\nleast one annotator has labeled it (Golbeck et al.,\n2017) or by a third objective instance (Gao and\nHuang, 2017). Most datasets use crowdsourcing\nplatforms or a few academic experts for annotation\n(Vidgen and Derczynski, 2020). Inter-annotatoragreement (IAA) and classification performance\nare established as two grounded evaluation measurements for annotation quality (Vidgen and Derczynski, 2020). Comparing the performance of amateur annotators (while providing guidelines) with\nexpert annotators for sexism and racism annotation,\nWaseem (2016) show that the quality of amateur\nannotators is competitive with expert annotations\nwhen several amateurs agree. Facing the trade-off\nbetween training annotators intensely and the number of involved annotators, we continued with the\ntrained annotators and group discussions/ individual revisions for flagged content and disagreements\n(Section 5.4).", "#### Who are the annotators?\n\n\n---|---\nGender|6 female, 2 male (8 total)\nAge:| 5 <30; 3 ≥30\nEthnicity:| 5 Danish: 1 Persian, 1 Arabic, 1 Polish\nStudy/occupation: | Linguistics (2); Health/Software Design; Ethnography/Digital Design; Communication/Psychology; Anthropology/Broadcast Moderator; Ethnography/Climate Change; Film Artist", "### Personal and Sensitive Information\n\n\nUsernames and PII were stripped during annotation process by skipping content containing these and eliding it from the final dataset\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe data contains abusive language. It may be possible to identify original speakers based on the content, so the data is only available for research purposes under a restrictive license and conditions. We hope that identifying sexism can help moderators. There is a possibility that the content here could be used to generate misogyny in Danish, which would place women in Denmark in an even more hostile environment, and for this reason data access is restricted and tracked.", "### Discussion of Biases\n\n\nWe have taken pains to mitigate as many biases as we were aware of in this work.\n\n\nSelection biases: Selection biases for abusive\nlanguage can be seen in the sampling of text, for instance when using keyword search (Wiegand et al.,\n2019), topic dependency (Ousidhoum et al., 2020), users (Wiegand et al., 2019), domain (Wiegand\net al., 2019), time (Florio et al., 2020) and lack of\nlinguistic variety (Vidgen and Derczynski, 2020).\n\n\nLabel biases: Label biases can be caused by, for\ninstance, non-representative annotator selection,\nlack in training/domain expertise, preconceived\nnotions, or pre-held stereotypes. These biases are\ntreated in relation to abusive language datasets\nby several sources, e.g. general sampling and\nannotators biases (Waseem, 2016; Al Kuwatly\net al., 2020), biases towards minority identity\nmentions based for example on gender or race\n(Davidson et al., 2017; Dixon et al., 2018; Park\net al., 2018; Davidson et al., 2019), and political\nannotator biases (Wich et al., 2020). Other qualitative biases comprise, for instance, demographic\nbias, over-generalization, topic exposure as social\nbiases (Hovy and Spruit, 2016).\n\n\nWe applied several measures to mitigate biases\noccurring through the annotation design and execution: First, we selected labels grounded in existing,\npeer-reviewed research from more than one field.\nSecond, we aimed for diversity in annotator profiles\nin terms of age, gender, dialect, and background.\nThird, we recruited a facilitator with a background\nin ethnographic studies and provided intense annotator training. Fourth, we engaged in weekly group\ndiscussions, iteratively improving the codebook\nand integrating edge cases. Fifth, the selection of\nplatforms from which we sampled data is based on\nlocal user representation in Denmark, rather than\nconvenience. Sixth, diverse sampling methods for\ndata collection reduced selection biases.", "### Other Known Limitations\n\n\nThe data is absolutely NOT a reasonable or in any way stratified sample of social media text, so class prevalence/balance here says nothing about incidences of these phenomena in the wild. That said, we hypothesis that the distribution of types of misogyny in this data (subtask C2) is roughly representative of how misogyny presents on the studied platforms.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors and the ethnographer-led annotation team.", "### Licensing Information\n\n\nThe data is licensed under a restrictive usage agreement. Apply for access here", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "language:da", "license:other", "not-for-all-audiences", "region:us" ]

[]

[ "da" ]

[ "### THIS PUBLIC-FACING DATASET IS A PREVIEW ONLY\n\n\nThis is a working data reader but the data here is just a preview of the full dataset, for safety & legal reasons.\n\n\nTo apply to access the entire dataset, complete this form.\n\n\nWhen you have the full data, amend '\\_URL' in 'URL' to point to the full data TSV's filename.", "### Dataset Summary\n\n\nThis is a high-quality dataset of annotated posts sampled from social\nmedia posts and annotated for misogyny. Danish language.\n\n\n\nSee the accompanying ACL paper Annotating Online Misogyny for full details.", "### Supported Tasks and Leaderboards\n\n\n* Hate Speech Detection on bajer\\_danish\\_misogyny", "### Languages\n\n\nDanish ('bcp47:da')\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### Bajer\n\n\nIn this preview: 10 instances\n\n\nIn the full dataset:\n\n\n* Size of downloaded dataset files: 7.29 MiB\n* Size of the generated dataset: 6.57 MiB\n* Total amount of disk used: 13.85 MiB\n\n\nSee above (or below) for how to get the full dataset.\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature, unique identifier in this dataset.\n* 'dataset\\_id': a 'string' feature, internal annotation identifier.\n* 'label\\_id': a 'string' feature, internal annotation sequence number.\n* 'text': a 'string' of the text that's annotated.\n* 'sampling': a 'string' describing which sampling technique surfaced this message\n* 'subtask\\_A': is the text abusive 'ABUS' or not 'NOT'? '0: NOT, 1: ABUS'\n* 'subtask\\_B': for abusive text, what's the target - individual 'IND', group 'GRP', other 'OTH', or untargeted 'UNT'? '0: IND, 1: GRP, 2: OTH, 3: UNT, 4: not applicable'\n* 'subtask\\_C1': for group-targeted abuse, what's the group - misogynistic 'SEX', other 'OTH', or racist 'RAC'? '0: SEX, 1: OTH, 2: RAC, 3: not applicable'\n* 'subtask\\_C2': for misogyny, is it neosexist 'NEOSEX', discrediting 'DISCREDIT', normative stereotyping 'NOR', benevolent sexism 'AMBIVALENT', dominance 'DOMINANCE', or harassment 'HARASSMENT'? '0: NEOSEX, 1: DISCREDIT, 2: NOR, 3: AMBIVALENT, 4: DOMINANCE, 5: HARASSMENT, 6: not applicable'", "### Data Splits\n\n\nIn the full dataset:\n\n\n\nThis preview has only 10 sentences - the link for access to the full data is given at the top of this page.\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nThe goal was to collect data for developing an annotation schema of online misogyny.\n\n\nRandom sampling of text often results in scarcity of examples of specifically misogynistic content (e.g. (Wulczyn et al., 2017;\nFounta et al., 2018)). Therefore, we used the common alternative of collecting data by using predefined keywords with a potentially high search hit\n(e.g. Waseem and Hovy (2016)), and identifying\nrelevant user-profiles (e.g. (Anzovino et al., 2018))\nand related topics (e.g. (Kumar et al., 2018)).\n\n\nWe searched for keywords (specific slurs, hashtags), that are known to occur in sexist posts. These\nwere defined by previous work, a slur list from\nReddit, and from interviews and surveys of online\nmisogyny among women. We also searched for\nbroader terms like “sex” or “women”, which do\nnot appear exclusively in a misogynistic context,\nfor example in the topic search, where we gathered\nrelevant posts and their comments from the social\nmedia pages of public media. A complete list of\nkeywords can be found in the appendix.\n\n\nSocial media provides a potentially biased, but\nbroad snapshot of online human discourse, with\nplenty of language and behaviours represented. Following best practice guidelines (Vidgen and Derczynski, 2020), we sampled from a language for\nwhich there are no existing annotations of the target\nphenomenon: Danish.\n\n\nDifferent social media platforms attract different user groups and can exhibit domain-specific\nlanguage (Karan and Snajder ˇ , 2018). Rather than\nchoosing one platform (existing misogyny datasets\nare primarily based on Twitter and Reddit (Guest\net al., 2021)), we sampled from multiple platforms:\nStatista (2020) shows that the platform where most\nDanish users are present is Facebook, followed\nby Twitter, YouTube, Instagram and lastly, Reddit.\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text. Data was gathered based on: keyword-based search (i.e. purposive sampling); topic-based search; and content from specific users.", "#### Who are the source language producers?\n\n\nDanish-speaking social media users", "### Annotations", "#### Annotation process\n\n\nIn annotating our dataset, we built on the MATTER\nframework (Pustejovsky and Stubbs, 2012) and use\nthe variation presented by Finlayson and Erjavec\n(2017) (the MALER framework), where the Train & Test stages are replaced by Leveraging of annotations for one’s particular goal, in our case the\ncreation of a comprehensive taxonomy.\n\n\nWe created a set of guidelines for the annotators.\nThe annotators were first asked to read the guidelines and individually annotate about 150 different\nposts, after which there was a shared discussion.\nAfter this pilot round, the volume of samples per annotator was increased and every sample labeled by\n2-3 annotators. When instances were ‘flagged’ or\nannotators disagreed on them, they were discussed\nduring weekly meetings, and misunderstandings\nwere resolved together with the external facilitator. After round three, when reaching 7k annotated\nposts (Figure 2), we continued with independent\nannotations maintaining a 15% instance overlap\nbetween randomly picked annotator pairs.\n\n\nManagement of annotator disagreement is an important part of the process design. Disagreements\ncan be solved by majority voting (Davidson et al.,\n2017; Wiegand et al., 2019), labeled as abuse if at\nleast one annotator has labeled it (Golbeck et al.,\n2017) or by a third objective instance (Gao and\nHuang, 2017). Most datasets use crowdsourcing\nplatforms or a few academic experts for annotation\n(Vidgen and Derczynski, 2020). Inter-annotatoragreement (IAA) and classification performance\nare established as two grounded evaluation measurements for annotation quality (Vidgen and Derczynski, 2020). Comparing the performance of amateur annotators (while providing guidelines) with\nexpert annotators for sexism and racism annotation,\nWaseem (2016) show that the quality of amateur\nannotators is competitive with expert annotations\nwhen several amateurs agree. Facing the trade-off\nbetween training annotators intensely and the number of involved annotators, we continued with the\ntrained annotators and group discussions/ individual revisions for flagged content and disagreements\n(Section 5.4).", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nUsernames and PII were stripped during annotation process by: skipping content containing these; and eliding it from the final dataset.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe data contains abusive language. It may be possible to identify original speakers based on the content, so the data is only available for research purposes under a restrictive license and conditions. We hope that identifying sexism can help moderators. There is a possibility that the content here could be used to generate misogyny in Danish, which would place women in Denmark in an even more hostile environment, and for this reason data access is restricted and tracked.", "### Discussion of Biases\n\n\nWe have taken pains to mitigate as many biases as we were aware of in this work.\n\n\nSelection biases: Selection biases for abusive\nlanguage can be seen in the sampling of text, for instance when using keyword search (Wiegand et al.,\n2019), topic dependency (Ousidhoum et al., 2020), users (Wiegand et al., 2019), domain (Wiegand\net al., 2019), time (Florio et al., 2020) and lack of\nlinguistic variety (Vidgen and Derczynski, 2020).\n\n\nLabel biases: Label biases can be caused by, for\ninstance, non-representative annotator selection,\nlack in training/domain expertise, preconceived\nnotions, or pre-held stereotypes. These biases are\ntreated in relation to abusive language datasets\nby several sources, e.g. general sampling and\nannotators biases (Waseem, 2016; Al Kuwatly\net al., 2020), biases towards minority identity\nmentions based for example on gender or race\n(Davidson et al., 2017; Dixon et al., 2018; Park\net al., 2018; Davidson et al., 2019), and political\nannotator biases (Wich et al., 2020). Other qualitative biases comprise, for instance, demographic\nbias, over-generalization, topic exposure as social\nbiases (Hovy and Spruit, 2016).\n\n\nWe applied several measures to mitigate biases\noccurring through the annotation design and execution: First, we selected labels grounded in existing,\npeer-reviewed research from more than one field.\nSecond, we aimed for diversity in annotator profiles\nin terms of age, gender, dialect, and background.\nThird, we recruited a facilitator with a background\nin ethnographic studies and provided intense annotator training. Fourth, we engaged in weekly group\ndiscussions, iteratively improving the codebook\nand integrating edge cases. Fifth, the selection of\nplatforms from which we sampled data is based on\nlocal user representation in Denmark, rather than\nconvenience. Sixth, diverse sampling methods for\ndata collection reduced selection biases.", "### Other Known Limitations\n\n\nThe data is absolutely NOT a reasonable or in any way stratified sample of social media text, so class prevalence/balance here says nothing about incidences of these phenomena in the wild. That said, we hypothesis that the distribution of types of misogyny in this data (subtask C2) is roughly representative of how misogyny presents on the studied platforms.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors and the ethnographer-led annotation team.", "### Licensing Information\n\n\nThe data is licensed under a restrictive usage agreement. Apply for access here", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-Danish #license-other #not-for-all-audiences #region-us \n", "### THIS PUBLIC-FACING DATASET IS A PREVIEW ONLY\n\n\nThis is a working data reader but the data here is just a preview of the full dataset, for safety & legal reasons.\n\n\nTo apply to access the entire dataset, complete this form.\n\n\nWhen you have the full data, amend '\\_URL' in 'URL' to point to the full data TSV's filename.", "### Dataset Summary\n\n\nThis is a high-quality dataset of annotated posts sampled from social\nmedia posts and annotated for misogyny. Danish language.\n\n\n\nSee the accompanying ACL paper Annotating Online Misogyny for full details.", "### Supported Tasks and Leaderboards\n\n\n* Hate Speech Detection on bajer\\_danish\\_misogyny", "### Languages\n\n\nDanish ('bcp47:da')\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### Bajer\n\n\nIn this preview: 10 instances\n\n\nIn the full dataset:\n\n\n* Size of downloaded dataset files: 7.29 MiB\n* Size of the generated dataset: 6.57 MiB\n* Total amount of disk used: 13.85 MiB\n\n\nSee above (or below) for how to get the full dataset.\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature, unique identifier in this dataset.\n* 'dataset\\_id': a 'string' feature, internal annotation identifier.\n* 'label\\_id': a 'string' feature, internal annotation sequence number.\n* 'text': a 'string' of the text that's annotated.\n* 'sampling': a 'string' describing which sampling technique surfaced this message\n* 'subtask\\_A': is the text abusive 'ABUS' or not 'NOT'? '0: NOT, 1: ABUS'\n* 'subtask\\_B': for abusive text, what's the target - individual 'IND', group 'GRP', other 'OTH', or untargeted 'UNT'? '0: IND, 1: GRP, 2: OTH, 3: UNT, 4: not applicable'\n* 'subtask\\_C1': for group-targeted abuse, what's the group - misogynistic 'SEX', other 'OTH', or racist 'RAC'? '0: SEX, 1: OTH, 2: RAC, 3: not applicable'\n* 'subtask\\_C2': for misogyny, is it neosexist 'NEOSEX', discrediting 'DISCREDIT', normative stereotyping 'NOR', benevolent sexism 'AMBIVALENT', dominance 'DOMINANCE', or harassment 'HARASSMENT'? '0: NEOSEX, 1: DISCREDIT, 2: NOR, 3: AMBIVALENT, 4: DOMINANCE, 5: HARASSMENT, 6: not applicable'", "### Data Splits\n\n\nIn the full dataset:\n\n\n\nThis preview has only 10 sentences - the link for access to the full data is given at the top of this page.\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nThe goal was to collect data for developing an annotation schema of online misogyny.\n\n\nRandom sampling of text often results in scarcity of examples of specifically misogynistic content (e.g. (Wulczyn et al., 2017;\nFounta et al., 2018)). Therefore, we used the common alternative of collecting data by using predefined keywords with a potentially high search hit\n(e.g. Waseem and Hovy (2016)), and identifying\nrelevant user-profiles (e.g. (Anzovino et al., 2018))\nand related topics (e.g. (Kumar et al., 2018)).\n\n\nWe searched for keywords (specific slurs, hashtags), that are known to occur in sexist posts. These\nwere defined by previous work, a slur list from\nReddit, and from interviews and surveys of online\nmisogyny among women. We also searched for\nbroader terms like “sex” or “women”, which do\nnot appear exclusively in a misogynistic context,\nfor example in the topic search, where we gathered\nrelevant posts and their comments from the social\nmedia pages of public media. A complete list of\nkeywords can be found in the appendix.\n\n\nSocial media provides a potentially biased, but\nbroad snapshot of online human discourse, with\nplenty of language and behaviours represented. Following best practice guidelines (Vidgen and Derczynski, 2020), we sampled from a language for\nwhich there are no existing annotations of the target\nphenomenon: Danish.\n\n\nDifferent social media platforms attract different user groups and can exhibit domain-specific\nlanguage (Karan and Snajder ˇ , 2018). Rather than\nchoosing one platform (existing misogyny datasets\nare primarily based on Twitter and Reddit (Guest\net al., 2021)), we sampled from multiple platforms:\nStatista (2020) shows that the platform where most\nDanish users are present is Facebook, followed\nby Twitter, YouTube, Instagram and lastly, Reddit.\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe dataset was sampled from Twitter, Facebook\nand Reddit posts as plain text. Data was gathered based on: keyword-based search (i.e. purposive sampling); topic-based search; and content from specific users.", "#### Who are the source language producers?\n\n\nDanish-speaking social media users", "### Annotations", "#### Annotation process\n\n\nIn annotating our dataset, we built on the MATTER\nframework (Pustejovsky and Stubbs, 2012) and use\nthe variation presented by Finlayson and Erjavec\n(2017) (the MALER framework), where the Train & Test stages are replaced by Leveraging of annotations for one’s particular goal, in our case the\ncreation of a comprehensive taxonomy.\n\n\nWe created a set of guidelines for the annotators.\nThe annotators were first asked to read the guidelines and individually annotate about 150 different\nposts, after which there was a shared discussion.\nAfter this pilot round, the volume of samples per annotator was increased and every sample labeled by\n2-3 annotators. When instances were ‘flagged’ or\nannotators disagreed on them, they were discussed\nduring weekly meetings, and misunderstandings\nwere resolved together with the external facilitator. After round three, when reaching 7k annotated\nposts (Figure 2), we continued with independent\nannotations maintaining a 15% instance overlap\nbetween randomly picked annotator pairs.\n\n\nManagement of annotator disagreement is an important part of the process design. Disagreements\ncan be solved by majority voting (Davidson et al.,\n2017; Wiegand et al., 2019), labeled as abuse if at\nleast one annotator has labeled it (Golbeck et al.,\n2017) or by a third objective instance (Gao and\nHuang, 2017). Most datasets use crowdsourcing\nplatforms or a few academic experts for annotation\n(Vidgen and Derczynski, 2020). Inter-annotatoragreement (IAA) and classification performance\nare established as two grounded evaluation measurements for annotation quality (Vidgen and Derczynski, 2020). Comparing the performance of amateur annotators (while providing guidelines) with\nexpert annotators for sexism and racism annotation,\nWaseem (2016) show that the quality of amateur\nannotators is competitive with expert annotations\nwhen several amateurs agree. Facing the trade-off\nbetween training annotators intensely and the number of involved annotators, we continued with the\ntrained annotators and group discussions/ individual revisions for flagged content and disagreements\n(Section 5.4).", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nUsernames and PII were stripped during annotation process by: skipping content containing these; and eliding it from the final dataset.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe data contains abusive language. It may be possible to identify original speakers based on the content, so the data is only available for research purposes under a restrictive license and conditions. We hope that identifying sexism can help moderators. There is a possibility that the content here could be used to generate misogyny in Danish, which would place women in Denmark in an even more hostile environment, and for this reason data access is restricted and tracked.", "### Discussion of Biases\n\n\nWe have taken pains to mitigate as many biases as we were aware of in this work.\n\n\nSelection biases: Selection biases for abusive\nlanguage can be seen in the sampling of text, for instance when using keyword search (Wiegand et al.,\n2019), topic dependency (Ousidhoum et al., 2020), users (Wiegand et al., 2019), domain (Wiegand\net al., 2019), time (Florio et al., 2020) and lack of\nlinguistic variety (Vidgen and Derczynski, 2020).\n\n\nLabel biases: Label biases can be caused by, for\ninstance, non-representative annotator selection,\nlack in training/domain expertise, preconceived\nnotions, or pre-held stereotypes. These biases are\ntreated in relation to abusive language datasets\nby several sources, e.g. general sampling and\nannotators biases (Waseem, 2016; Al Kuwatly\net al., 2020), biases towards minority identity\nmentions based for example on gender or race\n(Davidson et al., 2017; Dixon et al., 2018; Park\net al., 2018; Davidson et al., 2019), and political\nannotator biases (Wich et al., 2020). Other qualitative biases comprise, for instance, demographic\nbias, over-generalization, topic exposure as social\nbiases (Hovy and Spruit, 2016).\n\n\nWe applied several measures to mitigate biases\noccurring through the annotation design and execution: First, we selected labels grounded in existing,\npeer-reviewed research from more than one field.\nSecond, we aimed for diversity in annotator profiles\nin terms of age, gender, dialect, and background.\nThird, we recruited a facilitator with a background\nin ethnographic studies and provided intense annotator training. Fourth, we engaged in weekly group\ndiscussions, iteratively improving the codebook\nand integrating edge cases. Fifth, the selection of\nplatforms from which we sampled data is based on\nlocal user representation in Denmark, rather than\nconvenience. Sixth, diverse sampling methods for\ndata collection reduced selection biases.", "### Other Known Limitations\n\n\nThe data is absolutely NOT a reasonable or in any way stratified sample of social media text, so class prevalence/balance here says nothing about incidences of these phenomena in the wild. That said, we hypothesis that the distribution of types of misogyny in this data (subtask C2) is roughly representative of how misogyny presents on the studied platforms.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors and the ethnographer-led annotation team.", "### Licensing Information\n\n\nThe data is licensed under a restrictive usage agreement. Apply for access here", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "task_categories:text-generation", "task_ids:language-modeling", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:subjqa", "language:en", "license:cc-by-4.0", "question-generation", "arxiv:2210.03992", "region:us" ]

[ "2210.03992" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThis is a subset of QG-Bench, a unified question generation benchmark proposed in\n\"Generative Language Models for Paragraph-Level Question Generation: A Unified Benchmark and Evaluation, EMNLP 2022 main conference\".\nModified version of SubjQA for question generation (QG) task.", "### Supported Tasks and Leaderboards\n\n\n* 'question-generation': The dataset can be used to train a model for question generation.\nSuccess on this task is typically measured by achieving a high BLEU4/METEOR/ROUGE-L/BERTScore/MoverScore (see our paper for more in detail).", "### Languages\n\n\nEnglish (en)\n\n\nDataset Structure\n-----------------\n\n\nAn example of 'train' looks as follows.\n\n\nThe data fields are the same among all splits.\n\n\n* 'question': a 'string' feature.\n* 'paragraph': a 'string' feature.\n* 'answer': a 'string' feature.\n* 'sentence': a 'string' feature.\n* 'paragraph\\_answer': a 'string' feature, which is same as the paragraph but the answer is highlighted by a special token ''.\n* 'paragraph\\_sentence': a 'string' feature, which is same as the paragraph but a sentence containing the answer is highlighted by a special token ''.\n* 'sentence\\_answer': a 'string' feature, which is same as the sentence but the answer is highlighted by a special token ''.\n\n\nEach of 'paragraph\\_answer', 'paragraph\\_sentence', and 'sentence\\_answer' feature is assumed to be used to train a question generation model,\nbut with different information. The 'paragraph\\_answer' and 'sentence\\_answer' features are for answer-aware question generation and\n'paragraph\\_sentence' feature is for sentence-aware question generation.", "### Data Splits" ]

[ "TAGS\n#task_categories-text-generation #task_ids-language-modeling #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-subjqa #language-English #license-cc-by-4.0 #question-generation #arxiv-2210.03992 #region-us \n", "### Dataset Summary\n\n\nThis is a subset of QG-Bench, a unified question generation benchmark proposed in\n\"Generative Language Models for Paragraph-Level Question Generation: A Unified Benchmark and Evaluation, EMNLP 2022 main conference\".\nModified version of SubjQA for question generation (QG) task.", "### Supported Tasks and Leaderboards\n\n\n* 'question-generation': The dataset can be used to train a model for question generation.\nSuccess on this task is typically measured by achieving a high BLEU4/METEOR/ROUGE-L/BERTScore/MoverScore (see our paper for more in detail).", "### Languages\n\n\nEnglish (en)\n\n\nDataset Structure\n-----------------\n\n\nAn example of 'train' looks as follows.\n\n\nThe data fields are the same among all splits.\n\n\n* 'question': a 'string' feature.\n* 'paragraph': a 'string' feature.\n* 'answer': a 'string' feature.\n* 'sentence': a 'string' feature.\n* 'paragraph\\_answer': a 'string' feature, which is same as the paragraph but the answer is highlighted by a special token ''.\n* 'paragraph\\_sentence': a 'string' feature, which is same as the paragraph but a sentence containing the answer is highlighted by a special token ''.\n* 'sentence\\_answer': a 'string' feature, which is same as the sentence but the answer is highlighted by a special token ''.\n\n\nEach of 'paragraph\\_answer', 'paragraph\\_sentence', and 'sentence\\_answer' feature is assumed to be used to train a question generation model,\nbut with different information. The 'paragraph\\_answer' and 'sentence\\_answer' features are for answer-aware question generation and\n'paragraph\\_sentence' feature is for sentence-aware question generation.", "### Data Splits" ]

[ "region:us" ]

[]

[]

[ "## Information\nThis dataset shows 1785 manually annotated tweets from German politicians during the election year 2021 (01.01.2021 - 31.12.2021).\nThe tweets were annotated by 6 academics which were separated into two different groups. So every group of 3 people annotated the sentiment of ~900 tweets. For every tweet, the majority label was built. The annotation result had a moderate Kappa agreement.", "## Annotation\nThe tweets were annotated as follows:\n- 1 if the sentiment of the tweet is positive\n- 2 if the sentiment of the tweet is negative\n- 3 if the sentiment of the tweet is neutral" ]

[ "TAGS\n#region-us \n", "## Information\nThis dataset shows 1785 manually annotated tweets from German politicians during the election year 2021 (01.01.2021 - 31.12.2021).\nThe tweets were annotated by 6 academics which were separated into two different groups. So every group of 3 people annotated the sentiment of ~900 tweets. For every tweet, the majority label was built. The annotation result had a moderate Kappa agreement.", "## Annotation\nThe tweets were annotated as follows:\n- 1 if the sentiment of the tweet is positive\n- 2 if the sentiment of the tweet is negative\n- 3 if the sentiment of the tweet is neutral" ]

[ "task_categories:automatic-speech-recognition", "annotations_creators:expert-generated", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "language:en", "region:us" ]

[]

[ "en" ]

[ "### Dataset Summary\n\n\nThe TED-LIUM corpus is English-language TED talks, with transcriptions, sampled at 16kHz. The three releases of the corpus range from 118 to 452 hours of transcribed speech data.", "### Example", "### Supported Tasks and Leaderboards\n\n\n* 'automatic-speech-recognition': The dataset can be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER). The task has an active leaderboard which can be found at URL that ranks models based on their WER.", "### Languages\n\n\nThe audio and transcriptions are in English, as per the TED talks at URL.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* audio: A dictionary containing the path to the downloaded audio file, the decoded audio array, and the sampling rate. Note that when accessing the audio column: 'dataset[0][\"audio\"]' the audio file is automatically decoded and resampled to 'dataset.features[\"audio\"].sampling\\_rate'. Decoding and resampling of a large number of audio files might take a significant amount of time. Thus it is important to first query the sample index before the '\"audio\"' column, *i.e.* 'dataset[0][\"audio\"]' should always be preferred over 'dataset[\"audio\"][0]'.\n* file: A path to the downloaded audio file in .sph format.\n* text: the transcription of the audio file.\n* gender: the gender of the speaker. One of: male, female or N/A.\n* id: unique id of the data sample.\n* speaker\\_id: unique id of the speaker. The same speaker id can be found for multiple data samples.", "### Data Splits\n\n\nThere are three releases for the TED-LIUM corpus, progressively increasing the number of transcribed speech training data from 118 hours (Release 1), to 207 hours (Release 2), to 452 hours (Release 3).\n\n\nRelease 1:\n\n\n* 774 audio talks and automatically aligned transcriptions.\n* Contains 118 hours of speech audio data.\n* Homepage: URL\n\n\nRelease 2:\n\n\n* 1495 audio talks and automatically aligned transcriptions.\n* Contains 207 hours of speech audio data.\n* Dictionary with pronunciations (159848 entries).\n* Selected monolingual data for language modeling from WMT12 publicly available corpora.\n* Homepage: URL\n\n\nRelease 3:\n\n\n* 2351 audio talks and automatically aligned transcriptions.\n* Contains 452 hours of speech audio data.\n* TED-LIUM 2 validation and test data: 19 TED talks with their corresponding manual transcriptions.\n* Dictionary with pronunciations (159848 entries), the same file as the one included in TED-LIUM 2.\n* Selected monolingual data for language modeling from WMT12 publicly available corpora: these files come from the TED-LIUM 2 release, but have been modified to produce a tokenization more relevant for English language.\n* Homepage: URL\n\n\nRelease 3 contains two different corpus distributions:\n\n\n* The ‘legacy’ one, on which the dev and test datasets are the same as in TED-LIUM 2 (and TED-LIUM 1).\n* The ‘speaker adaptation’ one, specially designed for experiments on speaker adaptation.\n\n\nEach release is split into a training, validation and test set:\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nTED-LIUM was built during The International Workshop on Spoken Language Trans- lation (IWSLT) 2011 Evaluation Campaign, an annual workshop focused on the automatic translation of public talks and included tracks for speech recognition, speech translation, text translation, and system combination.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe data was obtained from publicly available TED talks at URL. Proper alignments between the speech and the transcribed text were generated using an in-house speaker segmentation and clustering tool (*LIUM\\_SpkDiarization*). Speech disfluencies (e.g. repetitions, hesitations, false starts) were treated in the following way: repetitions were transcribed, hesitations mapped to a specific filler word, and false starts not taken into account. For full details on the data collection and processing, refer to the TED-LIUM paper.", "#### Who are the source language producers?\n\n\nTED Talks are influential videos from expert speakers on education, business, science, tech and creativity.", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nLicensed under Creative Commons BY-NC-ND 3.0 (URL\n\n\nRelease 1:\n\n\nRelease 2:\n\n\nRelease 3:" ]

[ "TAGS\n#task_categories-automatic-speech-recognition #annotations_creators-expert-generated #language_creators-expert-generated #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-English #region-us \n", "### Dataset Summary\n\n\nThe TED-LIUM corpus is English-language TED talks, with transcriptions, sampled at 16kHz. The three releases of the corpus range from 118 to 452 hours of transcribed speech data.", "### Example", "### Supported Tasks and Leaderboards\n\n\n* 'automatic-speech-recognition': The dataset can be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER). The task has an active leaderboard which can be found at URL that ranks models based on their WER.", "### Languages\n\n\nThe audio and transcriptions are in English, as per the TED talks at URL.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* audio: A dictionary containing the path to the downloaded audio file, the decoded audio array, and the sampling rate. Note that when accessing the audio column: 'dataset[0][\"audio\"]' the audio file is automatically decoded and resampled to 'dataset.features[\"audio\"].sampling\\_rate'. Decoding and resampling of a large number of audio files might take a significant amount of time. Thus it is important to first query the sample index before the '\"audio\"' column, *i.e.* 'dataset[0][\"audio\"]' should always be preferred over 'dataset[\"audio\"][0]'.\n* file: A path to the downloaded audio file in .sph format.\n* text: the transcription of the audio file.\n* gender: the gender of the speaker. One of: male, female or N/A.\n* id: unique id of the data sample.\n* speaker\\_id: unique id of the speaker. The same speaker id can be found for multiple data samples.", "### Data Splits\n\n\nThere are three releases for the TED-LIUM corpus, progressively increasing the number of transcribed speech training data from 118 hours (Release 1), to 207 hours (Release 2), to 452 hours (Release 3).\n\n\nRelease 1:\n\n\n* 774 audio talks and automatically aligned transcriptions.\n* Contains 118 hours of speech audio data.\n* Homepage: URL\n\n\nRelease 2:\n\n\n* 1495 audio talks and automatically aligned transcriptions.\n* Contains 207 hours of speech audio data.\n* Dictionary with pronunciations (159848 entries).\n* Selected monolingual data for language modeling from WMT12 publicly available corpora.\n* Homepage: URL\n\n\nRelease 3:\n\n\n* 2351 audio talks and automatically aligned transcriptions.\n* Contains 452 hours of speech audio data.\n* TED-LIUM 2 validation and test data: 19 TED talks with their corresponding manual transcriptions.\n* Dictionary with pronunciations (159848 entries), the same file as the one included in TED-LIUM 2.\n* Selected monolingual data for language modeling from WMT12 publicly available corpora: these files come from the TED-LIUM 2 release, but have been modified to produce a tokenization more relevant for English language.\n* Homepage: URL\n\n\nRelease 3 contains two different corpus distributions:\n\n\n* The ‘legacy’ one, on which the dev and test datasets are the same as in TED-LIUM 2 (and TED-LIUM 1).\n* The ‘speaker adaptation’ one, specially designed for experiments on speaker adaptation.\n\n\nEach release is split into a training, validation and test set:\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nTED-LIUM was built during The International Workshop on Spoken Language Trans- lation (IWSLT) 2011 Evaluation Campaign, an annual workshop focused on the automatic translation of public talks and included tracks for speech recognition, speech translation, text translation, and system combination.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nThe data was obtained from publicly available TED talks at URL. Proper alignments between the speech and the transcribed text were generated using an in-house speaker segmentation and clustering tool (*LIUM\\_SpkDiarization*). Speech disfluencies (e.g. repetitions, hesitations, false starts) were treated in the following way: repetitions were transcribed, hesitations mapped to a specific filler word, and false starts not taken into account. For full details on the data collection and processing, refer to the TED-LIUM paper.", "#### Who are the source language producers?\n\n\nTED Talks are influential videos from expert speakers on education, business, science, tech and creativity.", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nLicensed under Creative Commons BY-NC-ND 3.0 (URL\n\n\nRelease 1:\n\n\nRelease 2:\n\n\nRelease 3:" ]

[ "task_categories:text-generation", "task_categories:fill-mask", "task_ids:language-modeling", "task_ids:masked-language-modeling", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:10M<n<100M", "source_datasets:original", "language:mt", "license:cc-by-nc-sa-4.0", "region:us" ]

[]

[ "mt" ]

[ "# Korpus Malti 🇲🇹\n\nGeneral Corpora for the Maltese Language.\n\nThis dataset is composed of texts from various genres/domains written in Maltese.", "## Configurations", "### Shuffled data\n\nThe default configuration ('\"shuffled\"') yields the the entire corpus from all genres:\n\n\nAll sentences are combined together and shuffled, without preserving the sentence order.\nNo other annotations are present, so an instance would be of the following form:\n\n\nThe training/validation/testing split is what was used to train the BERTu model.", "### Domain-split data\n\nAll other configurations contain a subset of the data.\nFor instance, this loads the Wikipedia portion:\n\n\nFor these configurations the data is not shuffled, so the sentence order on a document level is preserved.\nAn instance from these configurations would take the following form:\n\n\nThe raw data files contain additional metadata.\nIts structure differs from one instance to another, depending on what's available from the source.\nThis information was typically scraped from the source itself & minimal processing is performed on such data.", "## Additional Information", "### Dataset Curators\n\nThe dataset was created by Albert Gatt, Kurt Micallef, Marc Tanti, Lonneke van der Plas and Claudia Borg.", "### Licensing Information\n\nThis work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License][cc-by-nc-sa].\nPermissions beyond the scope of this license may be available at URL\n\n[![CC BY-NC-SA 4.0][cc-by-nc-sa-image]][cc-by-nc-sa]\n\n[cc-by-nc-sa]: URL\n[cc-by-nc-sa-image]: URL\n\n\n\nThis work was first presented in Pre-training Data Quality and Quantity for a Low-Resource Language: New Corpus and BERT Models for Maltese.\nCite it as follows:" ]

[ "TAGS\n#task_categories-text-generation #task_categories-fill-mask #task_ids-language-modeling #task_ids-masked-language-modeling #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-10M<n<100M #source_datasets-original #language-Maltese #license-cc-by-nc-sa-4.0 #region-us \n", "# Korpus Malti 🇲🇹\n\nGeneral Corpora for the Maltese Language.\n\nThis dataset is composed of texts from various genres/domains written in Maltese.", "## Configurations", "### Shuffled data\n\nThe default configuration ('\"shuffled\"') yields the the entire corpus from all genres:\n\n\nAll sentences are combined together and shuffled, without preserving the sentence order.\nNo other annotations are present, so an instance would be of the following form:\n\n\nThe training/validation/testing split is what was used to train the BERTu model.", "### Domain-split data\n\nAll other configurations contain a subset of the data.\nFor instance, this loads the Wikipedia portion:\n\n\nFor these configurations the data is not shuffled, so the sentence order on a document level is preserved.\nAn instance from these configurations would take the following form:\n\n\nThe raw data files contain additional metadata.\nIts structure differs from one instance to another, depending on what's available from the source.\nThis information was typically scraped from the source itself & minimal processing is performed on such data.", "## Additional Information", "### Dataset Curators\n\nThe dataset was created by Albert Gatt, Kurt Micallef, Marc Tanti, Lonneke van der Plas and Claudia Borg.", "### Licensing Information\n\nThis work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License][cc-by-nc-sa].\nPermissions beyond the scope of this license may be available at URL\n\n[![CC BY-NC-SA 4.0][cc-by-nc-sa-image]][cc-by-nc-sa]\n\n[cc-by-nc-sa]: URL\n[cc-by-nc-sa-image]: URL\n\n\n\nThis work was first presented in Pre-training Data Quality and Quantity for a Low-Resource Language: New Corpus and BERT Models for Maltese.\nCite it as follows:" ]

[ "region:us" ]

[]

[]

[ "# uzbek-sentiment-analysis\nSentiment analysis in the Uzbek language and new Datasets of Uzbek App reviews for Sentiment Classification\n\nFeel free to use the dataset and the tools presented in this project, a paper about more details on creation and usage here.\n\nIf you find it useful, please make sure to cite the paper:\n\n\n\nThe main contributions of this project are:\n\n1. The creation of the first annotated dataset for sentiment analysis in the Uzbek language, obtained from reviews of the top 100 Google Play Store applications used in Uzbekistan. This manually annotated dataset contains 2500 positive and 1800 negative reviews. Furthermore, we have also built a larger dataset by automatically translating (using Google Translate API) an existing English dataset of application reviews. The translated dataset has≈10K positive and≈10K negative app reviews, after manually eliminating the major machine translation errors by either correcting or removing them completely.\n\n2. The definition of the baselines for sentiment analyses in Uzbek by considering both traditional machine learning methods as well as recent deep learning techniques fed with fastText pre-trained word embeddings. Although all the tested models are relatively accurate and differences between models are small, the neural network models tested do not manage tosubstantially outperform traditional models. We believe that the quality of currently available pre-trained word embeddings for Uzbek is not enough to let deep learning models perform at their full potential.\n\n\nThe results obtained through the research:\n!Main Results Table\nTable: Accuracy results with different training and test sets.ManualTT- Manually annotated Training and Test sets.TransTT- Translated Training and Test sets.TTMT- Translated dataset for Training, Annotated dataset for Test set." ]

[ "TAGS\n#region-us \n", "# uzbek-sentiment-analysis\nSentiment analysis in the Uzbek language and new Datasets of Uzbek App reviews for Sentiment Classification\n\nFeel free to use the dataset and the tools presented in this project, a paper about more details on creation and usage here.\n\nIf you find it useful, please make sure to cite the paper:\n\n\n\nThe main contributions of this project are:\n\n1. The creation of the first annotated dataset for sentiment analysis in the Uzbek language, obtained from reviews of the top 100 Google Play Store applications used in Uzbekistan. This manually annotated dataset contains 2500 positive and 1800 negative reviews. Furthermore, we have also built a larger dataset by automatically translating (using Google Translate API) an existing English dataset of application reviews. The translated dataset has≈10K positive and≈10K negative app reviews, after manually eliminating the major machine translation errors by either correcting or removing them completely.\n\n2. The definition of the baselines for sentiment analyses in Uzbek by considering both traditional machine learning methods as well as recent deep learning techniques fed with fastText pre-trained word embeddings. Although all the tested models are relatively accurate and differences between models are small, the neural network models tested do not manage tosubstantially outperform traditional models. We believe that the quality of currently available pre-trained word embeddings for Uzbek is not enough to let deep learning models perform at their full potential.\n\n\nThe results obtained through the research:\n!Main Results Table\nTable: Accuracy results with different training and test sets.ManualTT- Manually annotated Training and Test sets.TransTT- Translated Training and Test sets.TTMT- Translated dataset for Training, Annotated dataset for Test set." ]

[ "region:us" ]

[]

[]

[ "### Dataset Summary\n\nGitHub Issues is a dataset consisting of GitHub issues and pull requests associated with the Datasets repository. It is intended for educational purposes and can be used for semantic search or multilabel text classification. The contents of each GitHub issue are in English and concern the domain of datasets for NLP, computer vision, and beyond.", "### Supported Tasks and Leaderboards\n\nFor each of the tasks tagged for this dataset, give a brief description of the tag, metrics, and suggested models (with a link to their HuggingFace implementation if available). Give a similar description of tasks that were not covered by the structured tag set (repace the 'task-category-tag' with an appropriate 'other:other-task-name').\n\n- 'task-category-tag': The dataset can be used to train a model for [TASK NAME], which consists in [TASK DESCRIPTION]. Success on this task is typically measured by achieving a *high/low* metric name. The (model name or model class) model currently achieves the following score. *[IF A LEADERBOARD IS AVAILABLE]:* This task has an active leaderboard which can be found at [leaderboard url]() and ranks models based on metric name while also reporting other metric name.", "### Languages\n\nProvide a brief overview of the languages represented in the dataset. Describe relevant details about specifics of the language such as whether it is social media text, African American English,...\n\nWhen relevant, please provide BCP-47 codes, which consist of a primary language subtag, with a script subtag and/or region subtag if available." ]

[ "TAGS\n#region-us \n", "### Dataset Summary\n\nGitHub Issues is a dataset consisting of GitHub issues and pull requests associated with the Datasets repository. It is intended for educational purposes and can be used for semantic search or multilabel text classification. The contents of each GitHub issue are in English and concern the domain of datasets for NLP, computer vision, and beyond.", "### Supported Tasks and Leaderboards\n\nFor each of the tasks tagged for this dataset, give a brief description of the tag, metrics, and suggested models (with a link to their HuggingFace implementation if available). Give a similar description of tasks that were not covered by the structured tag set (repace the 'task-category-tag' with an appropriate 'other:other-task-name').\n\n- 'task-category-tag': The dataset can be used to train a model for [TASK NAME], which consists in [TASK DESCRIPTION]. Success on this task is typically measured by achieving a *high/low* metric name. The (model name or model class) model currently achieves the following score. *[IF A LEADERBOARD IS AVAILABLE]:* This task has an active leaderboard which can be found at [leaderboard url]() and ranks models based on metric name while also reporting other metric name.", "### Languages\n\nProvide a brief overview of the languages represented in the dataset. Describe relevant details about specifics of the language such as whether it is social media text, African American English,...\n\nWhen relevant, please provide BCP-47 codes, which consist of a primary language subtag, with a script subtag and/or region subtag if available." ]

[ "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:1M<n<10M", "source_datasets:rendered|BookCorpusOpen", "language:en", "license:unknown", "arxiv:1506.06724", "arxiv:2207.06991", "arxiv:2105.05241", "region:us" ]

[ "1506.06724", "2207.06991", "2105.05241" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThis dataset is a version of the BookCorpus available at URL with examples rendered as images with resolution 16x8464 pixels.\n\n\nThe original BookCorpus was introduced by Zhu et al. (2015) in Aligning Books and Movies: Towards Story-Like Visual Explanations by Watching Movies and Reading Books and contains 17868 books of various genres. The rendered BookCorpus was used to train the PIXEL model introduced in the paper Language Modelling with Pixels by Phillip Rust, Jonas F. Lotz, Emanuele Bugliarello, Elizabeth Salesky, Miryam de Lhoneux, and Desmond Elliott.\n\n\nThe BookCorpusOpen dataset was rendered book-by-book into 5.4M examples containing approximately 1.1B words in total. The dataset is stored as a collection of 162 parquet files. It was rendered using the script openly available at URL The text renderer uses a PyGame backend and a collection of merged Google Noto Sans fonts. The PyGame backend does not support complex text layouts (e.g. ligatures and right-to-left scripts) or emoji, so occurrences of such text in the BookCorpus have not been rendered accurately.\nEach example consists of a \"pixel\\_values\" field which stores a 16x8464 (height, width) grayscale image containing the rendered text, and an integer value \"num\\_patches\" which stores how many image patches (when splitting the image into 529 non-overlapping patches of resolution 16x16 pixels) in the associated images contain actual text, i.e. are neither blank (fully white) nor are the fully black end-of-sequence patch.\n\n\nThe rendered BookCorpus can be loaded via the datasets library as follows:\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n* Size of downloaded dataset files: 63.58 GB\n* Size of the generated dataset: 63.59 GB\n* Total amount of disk used: 127.17 GB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.\n\n\n* 'pixel\\_values': an 'Image' feature.\n* 'num\\_patches': a 'Value(dtype=\"int64\")' feature.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThe books have been crawled from URL, see their terms of service for more information.\n\n\nA data sheet for this dataset has also been created and published in Addressing \"Documentation Debt\" in Machine Learning Research: A Retrospective Datasheet for BookCorpus", "### Contact Person\n\n\nThis dataset was added by Phillip Rust.\n\n\nGithub: @xplip\n\n\nTwitter: @rust\\_phillip" ]

[ "TAGS\n#annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-1M<n<10M #source_datasets-rendered|BookCorpusOpen #language-English #license-unknown #arxiv-1506.06724 #arxiv-2207.06991 #arxiv-2105.05241 #region-us \n", "### Dataset Summary\n\n\nThis dataset is a version of the BookCorpus available at URL with examples rendered as images with resolution 16x8464 pixels.\n\n\nThe original BookCorpus was introduced by Zhu et al. (2015) in Aligning Books and Movies: Towards Story-Like Visual Explanations by Watching Movies and Reading Books and contains 17868 books of various genres. The rendered BookCorpus was used to train the PIXEL model introduced in the paper Language Modelling with Pixels by Phillip Rust, Jonas F. Lotz, Emanuele Bugliarello, Elizabeth Salesky, Miryam de Lhoneux, and Desmond Elliott.\n\n\nThe BookCorpusOpen dataset was rendered book-by-book into 5.4M examples containing approximately 1.1B words in total. The dataset is stored as a collection of 162 parquet files. It was rendered using the script openly available at URL The text renderer uses a PyGame backend and a collection of merged Google Noto Sans fonts. The PyGame backend does not support complex text layouts (e.g. ligatures and right-to-left scripts) or emoji, so occurrences of such text in the BookCorpus have not been rendered accurately.\nEach example consists of a \"pixel\\_values\" field which stores a 16x8464 (height, width) grayscale image containing the rendered text, and an integer value \"num\\_patches\" which stores how many image patches (when splitting the image into 529 non-overlapping patches of resolution 16x16 pixels) in the associated images contain actual text, i.e. are neither blank (fully white) nor are the fully black end-of-sequence patch.\n\n\nThe rendered BookCorpus can be loaded via the datasets library as follows:\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n* Size of downloaded dataset files: 63.58 GB\n* Size of the generated dataset: 63.59 GB\n* Total amount of disk used: 127.17 GB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.\n\n\n* 'pixel\\_values': an 'Image' feature.\n* 'num\\_patches': a 'Value(dtype=\"int64\")' feature.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThe books have been crawled from URL, see their terms of service for more information.\n\n\nA data sheet for this dataset has also been created and published in Addressing \"Documentation Debt\" in Machine Learning Research: A Retrospective Datasheet for BookCorpus", "### Contact Person\n\n\nThis dataset was added by Phillip Rust.\n\n\nGithub: @xplip\n\n\nTwitter: @rust\\_phillip" ]

[ "annotations_creators:no-annotation", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:10M<n<100M", "source_datasets:original", "language:en", "license:cc-by-sa-3.0", "license:gfdl", "arxiv:2207.06991", "region:us" ]

[ "2207.06991" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThis dataset contains the full English Wikipedia from February 1, 2018, rendered into images of 16x8464 resolution.\n\n\nThe original text dataset was built from a Wikipedia dump. Each example in the original *text* dataset contained the content of one full Wikipedia article with cleaning to strip markdown and unwanted sections (references, etc.). Each *rendered* example contains a subset of one full article. This rendered English Wikipedia was used to train the PIXEL model introduced in the paper Language Modelling with Pixels by Phillip Rust, Jonas F. Lotz, Emanuele Bugliarello, Elizabeth Salesky, Miryam de Lhoneux, and Desmond Elliott.\n\n\nThe original Wikipedia text dataset was rendered article-by-article into 11.4M examples containing approximately 2B words in total. The dataset is stored as a collection of 338 parquet files.\n\n\nIt was rendered using the script openly available at URL The text renderer uses a PyGame backend and a collection of merged Google Noto Sans fonts. The PyGame backend does not support complex text layouts (e.g. ligatures and right-to-left scripts) or emoji, so occurrences of such text in the Wikipedia data have not been rendered accurately.\nEach example consists of a \"pixel\\_values\" field which stores a 16x8464 (height, width) grayscale image containing the rendered text, and an integer value \"num\\_patches\" which stores how many image patches (when splitting the image into 529 non-overlapping patches of resolution 16x16 pixels) in the associated images contain actual text, i.e. are neither blank (fully white) nor are the fully black end-of-sequence patch.\n\n\nYou can load the dataset as follows:\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n* Size of downloaded dataset files: 125.66 GB\n* Size of the generated dataset: 125.56 GB\n* Total amount of disk used: 251.22 GB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.\n\n\n* 'pixel\\_values': an 'Image' feature.\n* 'num\\_patches': a 'Value(dtype=\"int64\")' feature.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nMost of Wikipedia's text and many of its images are co-licensed under the Creative Commons Attribution-ShareAlike 3.0 Unported License (CC BY-SA) and the GNU Free Documentation License (GFDL) (unversioned, with no invariant sections, front-cover texts, or back-cover texts).\n\n\nSome text has been imported only under CC BY-SA and CC BY-SA-compatible license and cannot be reused under GFDL; such text will be identified on the page footer, in the page history, or on the discussion page of the article that utilizes the text.", "### Contact Person\n\n\nThis dataset was added by Phillip Rust.\n\n\nGithub: @xplip\n\n\nTwitter: @rust\\_phillip" ]

[ "TAGS\n#annotations_creators-no-annotation #language_creators-crowdsourced #multilinguality-monolingual #size_categories-10M<n<100M #source_datasets-original #language-English #license-cc-by-sa-3.0 #license-gfdl #arxiv-2207.06991 #region-us \n", "### Dataset Summary\n\n\nThis dataset contains the full English Wikipedia from February 1, 2018, rendered into images of 16x8464 resolution.\n\n\nThe original text dataset was built from a Wikipedia dump. Each example in the original *text* dataset contained the content of one full Wikipedia article with cleaning to strip markdown and unwanted sections (references, etc.). Each *rendered* example contains a subset of one full article. This rendered English Wikipedia was used to train the PIXEL model introduced in the paper Language Modelling with Pixels by Phillip Rust, Jonas F. Lotz, Emanuele Bugliarello, Elizabeth Salesky, Miryam de Lhoneux, and Desmond Elliott.\n\n\nThe original Wikipedia text dataset was rendered article-by-article into 11.4M examples containing approximately 2B words in total. The dataset is stored as a collection of 338 parquet files.\n\n\nIt was rendered using the script openly available at URL The text renderer uses a PyGame backend and a collection of merged Google Noto Sans fonts. The PyGame backend does not support complex text layouts (e.g. ligatures and right-to-left scripts) or emoji, so occurrences of such text in the Wikipedia data have not been rendered accurately.\nEach example consists of a \"pixel\\_values\" field which stores a 16x8464 (height, width) grayscale image containing the rendered text, and an integer value \"num\\_patches\" which stores how many image patches (when splitting the image into 529 non-overlapping patches of resolution 16x16 pixels) in the associated images contain actual text, i.e. are neither blank (fully white) nor are the fully black end-of-sequence patch.\n\n\nYou can load the dataset as follows:\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n* Size of downloaded dataset files: 125.66 GB\n* Size of the generated dataset: 125.56 GB\n* Total amount of disk used: 251.22 GB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.\n\n\n* 'pixel\\_values': an 'Image' feature.\n* 'num\\_patches': a 'Value(dtype=\"int64\")' feature.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nMost of Wikipedia's text and many of its images are co-licensed under the Creative Commons Attribution-ShareAlike 3.0 Unported License (CC BY-SA) and the GNU Free Documentation License (GFDL) (unversioned, with no invariant sections, front-cover texts, or back-cover texts).\n\n\nSome text has been imported only under CC BY-SA and CC BY-SA-compatible license and cannot be reused under GFDL; such text will be identified on the page footer, in the page history, or on the discussion page of the article that utilizes the text.", "### Contact Person\n\n\nThis dataset was added by Phillip Rust.\n\n\nGithub: @xplip\n\n\nTwitter: @rust\\_phillip" ]

[ "task_categories:token-classification", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:cc-by-4.0", "region:us" ]

[]

[ "en" ]

[ "### Dataset Summary\n\n\nThis is a dataset annotated for *named temporal expression* chunks.\n\n\nThe\ncommonest temporal expressions typically\ncontain date and time words, like April or\nhours. Research into recognising and interpreting these typical expressions is mature in many languages. However, there is\na class of expressions that are less typical,\nvery varied, and difficult to automatically\ninterpret. These indicate dates and times,\nbut are harder to detect because they often do not contain time words and are not\nused frequently enough to appear in conventional temporally-annotated corpora –\nfor example *Michaelmas* or *Vasant Panchami*.\n\n\nFor more details see Recognising and Interpreting Named Temporal Expressions", "### Supported Tasks and Leaderboards\n\n\n* Task: Named Entity Recognition (temporal expressions)", "### Languages\n\n\nEnglsih\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\nEach tweet contains an ID, a list of tokens, and a list of timex chunk flags.\n\n\n* 'id': a 'string' feature.\n* 'tokens': a 'list' of 'strings' .\n* 'ntimex\\_tags': a 'list' of class IDs ('int's) for whether a token is out-of-timex or in a timex chunk.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nCreative Commons Attribution 4.0 International (CC BY 4.0)", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "TAGS\n#task_categories-token-classification #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-cc-by-4.0 #region-us \n", "### Dataset Summary\n\n\nThis is a dataset annotated for *named temporal expression* chunks.\n\n\nThe\ncommonest temporal expressions typically\ncontain date and time words, like April or\nhours. Research into recognising and interpreting these typical expressions is mature in many languages. However, there is\na class of expressions that are less typical,\nvery varied, and difficult to automatically\ninterpret. These indicate dates and times,\nbut are harder to detect because they often do not contain time words and are not\nused frequently enough to appear in conventional temporally-annotated corpora –\nfor example *Michaelmas* or *Vasant Panchami*.\n\n\nFor more details see Recognising and Interpreting Named Temporal Expressions", "### Supported Tasks and Leaderboards\n\n\n* Task: Named Entity Recognition (temporal expressions)", "### Languages\n\n\nEnglsih\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\nEach tweet contains an ID, a list of tokens, and a list of timex chunk flags.\n\n\n* 'id': a 'string' feature.\n* 'tokens': a 'list' of 'strings' .\n* 'ntimex\\_tags': a 'list' of class IDs ('int's) for whether a token is out-of-timex or in a timex chunk.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nCreative Commons Attribution 4.0 International (CC BY 4.0)", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "task_categories:translation", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:multilingual", "size_categories:1K<n<10K", "source_datasets:original", "language:da", "language:fo", "license:cc-by-4.0", "arxiv:2206.08727", "doi:10.57967/hf/0515", "region:us" ]

[ "2206.08727" ]

[ "da", "fo" ]

[ "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: URL\n- Leaderboard: \n- Point of Contact: Leon Derczynski", "### Dataset Summary\n\nThis is a native-speaker-generated parallel corpus of Faroese and Danish", "### Supported Tasks and Leaderboards\n\n*", "### Languages\n\n* Danish\n* Faroese", "## Dataset Structure", "### Data Instances\n3995 parallel sentences", "### Data Fields\n\n* 'id': the sentence pair ID, 'string'\n* 'origin': the original sentence identifier text, 'string'\n* 'fo': the Faroese text, 'string'\n* 'da': the Danish text, 'string'", "### Data Splits\n\nMonolithic", "## Dataset Creation", "### Curation Rationale\n\nTo gather a broad range of topics about the Faroes and the rest of the world, to enable a general-purpose Faroese:Danish translation system", "### Source Data", "#### Initial Data Collection and Normalization\n\n* EUROparl Danish\n* Dimmaletting, Faroese newspaper\n* Tatoeba Danish / Faroese", "#### Who are the source language producers?", "### Annotations", "#### Annotation process\n\nNo annotations", "#### Who are the annotators?\n\nTwo Faroese native speakers, one male one female, in their 20s, masters degrees, living in Denmark", "### Personal and Sensitive Information\n\nNone due to the sources used", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis collection of Faroese is curated by Leon Derczynski", "### Licensing Information\n\nCreative Commons Attribution 4.0" ]

[ "TAGS\n#task_categories-translation #annotations_creators-expert-generated #language_creators-found #multilinguality-multilingual #size_categories-1K<n<10K #source_datasets-original #language-Danish #language-Faroese #license-cc-by-4.0 #arxiv-2206.08727 #doi-10.57967/hf/0515 #region-us \n", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: URL\n- Leaderboard: \n- Point of Contact: Leon Derczynski", "### Dataset Summary\n\nThis is a native-speaker-generated parallel corpus of Faroese and Danish", "### Supported Tasks and Leaderboards\n\n*", "### Languages\n\n* Danish\n* Faroese", "## Dataset Structure", "### Data Instances\n3995 parallel sentences", "### Data Fields\n\n* 'id': the sentence pair ID, 'string'\n* 'origin': the original sentence identifier text, 'string'\n* 'fo': the Faroese text, 'string'\n* 'da': the Danish text, 'string'", "### Data Splits\n\nMonolithic", "## Dataset Creation", "### Curation Rationale\n\nTo gather a broad range of topics about the Faroes and the rest of the world, to enable a general-purpose Faroese:Danish translation system", "### Source Data", "#### Initial Data Collection and Normalization\n\n* EUROparl Danish\n* Dimmaletting, Faroese newspaper\n* Tatoeba Danish / Faroese", "#### Who are the source language producers?", "### Annotations", "#### Annotation process\n\nNo annotations", "#### Who are the annotators?\n\nTwo Faroese native speakers, one male one female, in their 20s, masters degrees, living in Denmark", "### Personal and Sensitive Information\n\nNone due to the sources used", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis collection of Faroese is curated by Leon Derczynski", "### Licensing Information\n\nCreative Commons Attribution 4.0" ]

[ "language:ar", "region:us" ]

[]

[ "ar" ]

[ "# Flickr8k Image Features\n\nFlickr8k image features are extracted using the ResNeXt-152 C4 architecture (found here) and can be used as input for the OSCAR learning method. Arabic captions and splits are provided by ElJundi et al.", "## Dev-split\n+ URL Yaml configure file with Arabic object tags\n+ URL Extracted image features\n+ URL Arabic labels\n+ URL English labels\n+ URL Yaml configure file with English object tags\n+ dev_caption.json Arabic captions for training\n+ dev_caption_coco_format.json Arabic captions for validation", "## Test-split\n+ URL Yaml configure file with Arabic object tags\n+ URL Extracted image features\n+ URL Arabic labels\n+ URL English labels\n+ URL Yaml configure file with English object tags\n+ test_caption.json Arabic captions for training\n+ test_caption_coco_format.json Arabic captions for validation", "## Train-split\n+ URL Yaml configure file with Arabic object tags\n+ URL Extracted image features\n+ URL Arabic labels\n+ URL English labels\n+ URL Yaml configure file with English object tags\n+ train_caption.json Arabic captions for training\n+ train_caption_coco_format.json Arabic captions for validation" ]

[ "TAGS\n#language-Arabic #region-us \n", "# Flickr8k Image Features\n\nFlickr8k image features are extracted using the ResNeXt-152 C4 architecture (found here) and can be used as input for the OSCAR learning method. Arabic captions and splits are provided by ElJundi et al.", "## Dev-split\n+ URL Yaml configure file with Arabic object tags\n+ URL Extracted image features\n+ URL Arabic labels\n+ URL English labels\n+ URL Yaml configure file with English object tags\n+ dev_caption.json Arabic captions for training\n+ dev_caption_coco_format.json Arabic captions for validation", "## Test-split\n+ URL Yaml configure file with Arabic object tags\n+ URL Extracted image features\n+ URL Arabic labels\n+ URL English labels\n+ URL Yaml configure file with English object tags\n+ test_caption.json Arabic captions for training\n+ test_caption_coco_format.json Arabic captions for validation", "## Train-split\n+ URL Yaml configure file with Arabic object tags\n+ URL Extracted image features\n+ URL Arabic labels\n+ URL English labels\n+ URL Yaml configure file with English object tags\n+ train_caption.json Arabic captions for training\n+ train_caption_coco_format.json Arabic captions for validation" ]

[ "task_categories:question-answering", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "annotations_creators:expert-generated", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:n<1K", "source_datasets:original", "language:en", "license:apache-2.0", "arxiv:2109.07958", "region:us" ]

[ "2109.07958" ]

[ "en" ]

[ "# Dataset Card for TruthfulQA", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: URL", "### Dataset Summary\n\nTruthfulQA: Measuring How Models Mimic Human Falsehoods\n\nWe propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web.", "### Supported Tasks and Leaderboards\n\nSee: Tasks", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nThe benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics.", "### Data Fields\n\n1. Type: Adversarial v Non-Adversarial Questions\n2. Category: Category of misleading question\n3. Question: The question\n4. Best Answer: The best correct answer\n5. Correct Answers: A set of correct answers. Delimited by ';'.\n6. Incorrect Answers: A set of incorrect answers. Delimited by ';'.\n7. Source: A source that supports the correct answers.", "### Data Splits\n\nDue to constraints of huggingface the dataset is loaded into a \"train\" split.", "### Contributions\n\nThanks to @sylinrl for adding this dataset." ]

[ "TAGS\n#task_categories-question-answering #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #annotations_creators-expert-generated #language_creators-expert-generated #multilinguality-monolingual #size_categories-n<1K #source_datasets-original #language-English #license-apache-2.0 #arxiv-2109.07958 #region-us \n", "# Dataset Card for TruthfulQA", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: URL", "### Dataset Summary\n\nTruthfulQA: Measuring How Models Mimic Human Falsehoods\n\nWe propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web.", "### Supported Tasks and Leaderboards\n\nSee: Tasks", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nThe benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics.", "### Data Fields\n\n1. Type: Adversarial v Non-Adversarial Questions\n2. Category: Category of misleading question\n3. Question: The question\n4. Best Answer: The best correct answer\n5. Correct Answers: A set of correct answers. Delimited by ';'.\n6. Incorrect Answers: A set of incorrect answers. Delimited by ';'.\n7. Source: A source that supports the correct answers.", "### Data Splits\n\nDue to constraints of huggingface the dataset is loaded into a \"train\" split.", "### Contributions\n\nThanks to @sylinrl for adding this dataset." ]

[ "task_categories:token-classification", "task_ids:named-entity-recognition", "annotations_creators:machine-generated", "annotations_creators:expert-generated", "language_creators:found", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:other", "epidemiology", "rare disease", "named entity recognition", "NER", "NIH", "region:us" ]

[]

[ "en" ]

[ "### Dataset Summary\n\n\nEpiSet4NER-v2 is a gold-standard dataset for epidemiological entity recognition of location, epidemiologic types (e.g. \"prevalence\", \"annual incidence\", \"estimated occurrence\"), and epidemiological rates (e.g. \"1.7 per 1,000,000 live births\", \"2.1:1.000.000\", \"one in five million\", \"0.03%\") created by the Genetic and Rare Diseases Information Center (GARD), a program in the National Center for Advancing Translational Sciences, one of the 27 National Institutes of Health. It was labeled programmatically using spaCy NER and rule-based methods, then manually validated by biomedical researchers, including a GARD curator (genetic and rare disease expert). This weakly-supervised teaching method allowed us to construct this high quality dataset in an efficient manner and achieve satisfactory performance on a multi-type token classification problem. It was used to train EpiExtract4GARD-v2, a BioBERT-based model fine-tuned for NER.", "### Data Fields\n\n\nThe data fields are the same among all splits.\n\n\n* 'id': a 'string' feature that indicates sentence number.\n* 'tokens': a 'list' of 'string' features.\n* 'ner\\_tags': a 'list' of classification labels, with possible values including 'O' (0), 'B-LOC' (1), 'I-LOC' (2), 'B-EPI' (3), 'I-EPI' (4),'B-STAT' (5),'I-STAT' (6).", "### Data Splits\n\n\n\nDataset Creation\n----------------\n\n\n!EpiSet Creation Flowchart\n*Figure 1:* Creation of EpiSet4NER by NIH/NCATS\nComparing the programmatically labeled test set to the manually corrected test set allowed us to measure the precision, recall, and F1 of the programmatic labeling.\n\n\n*Table 1:* Programmatic labeling of EpiSet4NER\n\n\n\nAn example of the text labeling:\n!Text Labeling\n*Figure 2:* Text Labeling using spaCy and rule-based labeling. Ideal labeling is bolded on the left. Actual programmatic output is on the right. [[Figure citation]](URL", "### Curation Rationale\n\n\nTo train ML/DL models that automate the process of rare disease epidemiological curation. This is crucial information to patients & families, researchers, grantors, and policy makers, primarily for funding purposes.", "### Source Data\n\n\n620 rare disease abstracts classified as epidemiological by a LSTM RNN rare disease epi classifier from 488 diseases. See Figure 1.", "#### Initial Data Collection and Normalization\n\n\nA random sample of 500 disease names were gathered from a list of ~6061 rare diseases tracked by GARD until ≥50 abstracts had been returned for each disease or the EBI RESTful API results were exhausted. Though we called ~25,000 abstracts from PubMed's db, only 7699 unique abstracts were returned for 488 diseases. Out of 7699 abstracts, only 620 were classified as epidemiological by the LSTM RNN epidemiological classifier.", "### Annotations", "#### Annotation process\n\n\nProgrammatic labeling. See here and then here. The test set was manually corrected after creation.", "#### Who are the annotators?\n\n\nProgrammatic labeling was done by @William Kariampuzha, one of the NCATS researchers.\nThe test set was manually corrected by 2 more NCATS researchers and a GARD curator (genetic and rare disease expert).", "### Personal and Sensitive Information\n\n\nNone. These are freely available abstracts from PubMed.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nAssisting 25-30 millions Americans with rare diseases. Additionally can be useful for Orphanet or CDC researchers/curators.", "### Discussion of Biases and Limitations\n\n\n* There were errors in the source file that contained rare disease synonyms of names, which may have led to some unrelated abstracts being included in the training, validation, and test sets.\n* The abstracts were gathered through the EBI API and is thus subject to any biases that the EBI API had. The NCBI API returns very different results as shown by an API analysis here.\n* The long short-term memory recurrent neural network epi classifier was used to sift the 7699 rare disease abstracts. This model had a hold-out validation F1 score of 0.886 and a test F1 (which was compared against a GARD curator who used full-text articles to determine truth-value of epidemiological abstract) of 0.701. With 620 epi abstracts filtered from 7699 original rare disease abstracts, there are likely several false positives and false negative epi abstracts.\n* Tokenization was done by spaCy which may be a limitation (or not) for current and future models trained on this set.\n* The programmatic labeling was very imprecise as seen by Table 1. This is likely the largest limitation of the BioBERT-based model trained on this set.\n* The test set was difficult to validate even for general NCATS researchers, which is why we relied on a rare disease expert to verify our modifications. As this task of epidemiological information identification is quite difficult for non-expert humans to complete, this set, and especially a gold-standard dataset in the possible future, represents a challenging gauntlet for NLP systems, especially those focusing on numeracy, to compete on.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nNIH GARD", "### Licensing Information", "### Contributions\n\n\nThanks to @William Kariampuzha at NCATS/Axle Informatics for adding this dataset." ]

[ "TAGS\n#task_categories-token-classification #task_ids-named-entity-recognition #annotations_creators-machine-generated #annotations_creators-expert-generated #language_creators-found #language_creators-expert-generated #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-other #epidemiology #rare disease #named entity recognition #NER #NIH #region-us \n", "### Dataset Summary\n\n\nEpiSet4NER-v2 is a gold-standard dataset for epidemiological entity recognition of location, epidemiologic types (e.g. \"prevalence\", \"annual incidence\", \"estimated occurrence\"), and epidemiological rates (e.g. \"1.7 per 1,000,000 live births\", \"2.1:1.000.000\", \"one in five million\", \"0.03%\") created by the Genetic and Rare Diseases Information Center (GARD), a program in the National Center for Advancing Translational Sciences, one of the 27 National Institutes of Health. It was labeled programmatically using spaCy NER and rule-based methods, then manually validated by biomedical researchers, including a GARD curator (genetic and rare disease expert). This weakly-supervised teaching method allowed us to construct this high quality dataset in an efficient manner and achieve satisfactory performance on a multi-type token classification problem. It was used to train EpiExtract4GARD-v2, a BioBERT-based model fine-tuned for NER.", "### Data Fields\n\n\nThe data fields are the same among all splits.\n\n\n* 'id': a 'string' feature that indicates sentence number.\n* 'tokens': a 'list' of 'string' features.\n* 'ner\\_tags': a 'list' of classification labels, with possible values including 'O' (0), 'B-LOC' (1), 'I-LOC' (2), 'B-EPI' (3), 'I-EPI' (4),'B-STAT' (5),'I-STAT' (6).", "### Data Splits\n\n\n\nDataset Creation\n----------------\n\n\n!EpiSet Creation Flowchart\n*Figure 1:* Creation of EpiSet4NER by NIH/NCATS\nComparing the programmatically labeled test set to the manually corrected test set allowed us to measure the precision, recall, and F1 of the programmatic labeling.\n\n\n*Table 1:* Programmatic labeling of EpiSet4NER\n\n\n\nAn example of the text labeling:\n!Text Labeling\n*Figure 2:* Text Labeling using spaCy and rule-based labeling. Ideal labeling is bolded on the left. Actual programmatic output is on the right. [[Figure citation]](URL", "### Curation Rationale\n\n\nTo train ML/DL models that automate the process of rare disease epidemiological curation. This is crucial information to patients & families, researchers, grantors, and policy makers, primarily for funding purposes.", "### Source Data\n\n\n620 rare disease abstracts classified as epidemiological by a LSTM RNN rare disease epi classifier from 488 diseases. See Figure 1.", "#### Initial Data Collection and Normalization\n\n\nA random sample of 500 disease names were gathered from a list of ~6061 rare diseases tracked by GARD until ≥50 abstracts had been returned for each disease or the EBI RESTful API results were exhausted. Though we called ~25,000 abstracts from PubMed's db, only 7699 unique abstracts were returned for 488 diseases. Out of 7699 abstracts, only 620 were classified as epidemiological by the LSTM RNN epidemiological classifier.", "### Annotations", "#### Annotation process\n\n\nProgrammatic labeling. See here and then here. The test set was manually corrected after creation.", "#### Who are the annotators?\n\n\nProgrammatic labeling was done by @William Kariampuzha, one of the NCATS researchers.\nThe test set was manually corrected by 2 more NCATS researchers and a GARD curator (genetic and rare disease expert).", "### Personal and Sensitive Information\n\n\nNone. These are freely available abstracts from PubMed.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nAssisting 25-30 millions Americans with rare diseases. Additionally can be useful for Orphanet or CDC researchers/curators.", "### Discussion of Biases and Limitations\n\n\n* There were errors in the source file that contained rare disease synonyms of names, which may have led to some unrelated abstracts being included in the training, validation, and test sets.\n* The abstracts were gathered through the EBI API and is thus subject to any biases that the EBI API had. The NCBI API returns very different results as shown by an API analysis here.\n* The long short-term memory recurrent neural network epi classifier was used to sift the 7699 rare disease abstracts. This model had a hold-out validation F1 score of 0.886 and a test F1 (which was compared against a GARD curator who used full-text articles to determine truth-value of epidemiological abstract) of 0.701. With 620 epi abstracts filtered from 7699 original rare disease abstracts, there are likely several false positives and false negative epi abstracts.\n* Tokenization was done by spaCy which may be a limitation (or not) for current and future models trained on this set.\n* The programmatic labeling was very imprecise as seen by Table 1. This is likely the largest limitation of the BioBERT-based model trained on this set.\n* The test set was difficult to validate even for general NCATS researchers, which is why we relied on a rare disease expert to verify our modifications. As this task of epidemiological information identification is quite difficult for non-expert humans to complete, this set, and especially a gold-standard dataset in the possible future, represents a challenging gauntlet for NLP systems, especially those focusing on numeracy, to compete on.\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nNIH GARD", "### Licensing Information", "### Contributions\n\n\nThanks to @William Kariampuzha at NCATS/Axle Informatics for adding this dataset." ]

[ "task_categories:text-classification", "task_ids:fact-checking", "annotations_creators:crowdsourced", "language_creators:found", "multilinguality:monolingual", "size_categories:10K<n<100K", "language:en", "license:cc-by-4.0", "stance-detection", "arxiv:1809.06683", "region:us" ]

[ "1809.06683" ]

[ "en" ]

[ "### Dataset Summary\n\n\nStance prediction task in English. The goal is to predict whether a given reply to a claim either supports, denies, questions, or simply comments on the claim. Ran as a SemEval task in 2019.", "### Supported Tasks and Leaderboards\n\n\n* SemEval 2019 task 1", "### Languages\n\n\nEnglish of various origins, bcp47: 'en'\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### polstance\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature.\n* 'source\\_text': a 'string' expressing a claim/topic.\n* 'reply\\_text': a 'string' to be classified for its stance to the source.\n* 'label': a class label representing the stance the text expresses towards the target. Full tagset with indices:\n* 'quoteID': a 'string' of the internal quote ID.\n* 'party': a 'string' describing the party affiliation of the quote utterer at the time of utterance.\n* 'politician': a 'string' naming the politician who uttered the quote.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?\n\n\nTwitter users", "### Annotations", "#### Annotation process\n\n\nDetailed in Analysing How People Orient to and Spread Rumours in Social Media by Looking at Conversational Threads", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors.", "### Licensing Information\n\n\nThe authors distribute this data under Creative Commons attribution license, CC-BY 4.0.", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "TAGS\n#task_categories-text-classification #task_ids-fact-checking #annotations_creators-crowdsourced #language_creators-found #multilinguality-monolingual #size_categories-10K<n<100K #language-English #license-cc-by-4.0 #stance-detection #arxiv-1809.06683 #region-us \n", "### Dataset Summary\n\n\nStance prediction task in English. The goal is to predict whether a given reply to a claim either supports, denies, questions, or simply comments on the claim. Ran as a SemEval task in 2019.", "### Supported Tasks and Leaderboards\n\n\n* SemEval 2019 task 1", "### Languages\n\n\nEnglish of various origins, bcp47: 'en'\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### polstance\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\n* 'id': a 'string' feature.\n* 'source\\_text': a 'string' expressing a claim/topic.\n* 'reply\\_text': a 'string' to be classified for its stance to the source.\n* 'label': a class label representing the stance the text expresses towards the target. Full tagset with indices:\n* 'quoteID': a 'string' of the internal quote ID.\n* 'party': a 'string' describing the party affiliation of the quote utterer at the time of utterance.\n* 'politician': a 'string' naming the politician who uttered the quote.", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?\n\n\nTwitter users", "### Annotations", "#### Annotation process\n\n\nDetailed in Analysing How People Orient to and Spread Rumours in Social Media by Looking at Conversational Threads", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThe dataset is curated by the paper's authors.", "### Licensing Information\n\n\nThe authors distribute this data under Creative Commons attribution license, CC-BY 4.0.", "### Contributions\n\n\nAuthor-added dataset @leondz" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "license:ecl-2.0", "region:us" ]

[]

[]

[ "# Context\nThis dataset contains all the stats of all club goals of Cristiano Ronaldo dos Santos Aveiro.", "# About Cristiano Ronaldo\nCristiano Ronaldo dos Santos Aveiro is a Portuguese professional footballer who plays as a forward for Premier League club Manchester United and captains the Portugal national team.\n \n- Current team: Portugal national football team (#7 / Forward) Trending\n \n- Born: February 5, 1985 (age 37 years), Hospital Dr. Nélio Mendonça, Funchal, Portugal\n- Height: 1.87 m\n- Partner: Georgina Rodríguez (2017–)\n- Salary: 26.52 million GBP (2022)\n- Children: Cristiano Ronaldo Jr., Alana Martina dos Santos Aveiro, Eva Maria Dos Santos, Mateo Ronaldo\n\n!CR7", "# Content\n- URL file containing Goal_no, Season, Competition, Matchday, Venue, Team, Opponent, Result, Position, Minute, At_score, Type_of_goal", "# Featured Notebook\nCR7 - Extensive EDA & Analytics-Cristiano Ronaldo", "# GitHub Project\n- Data Collection : GitHub", "# Download\nkaggle API Command\n\n'!kaggle datasets download -d azminetoushikwasi/cr7-cristiano-ronaldo-all-club-goals-stats'", "## Disclaimer\nThe data collected are all publicly available and it's intended for educational purposes only.", "## Acknowledgement\nCover image credit - URL" ]

[ "TAGS\n#license-ecl-2.0 #region-us \n", "# Context\nThis dataset contains all the stats of all club goals of Cristiano Ronaldo dos Santos Aveiro.", "# About Cristiano Ronaldo\nCristiano Ronaldo dos Santos Aveiro is a Portuguese professional footballer who plays as a forward for Premier League club Manchester United and captains the Portugal national team.\n \n- Current team: Portugal national football team (#7 / Forward) Trending\n \n- Born: February 5, 1985 (age 37 years), Hospital Dr. Nélio Mendonça, Funchal, Portugal\n- Height: 1.87 m\n- Partner: Georgina Rodríguez (2017–)\n- Salary: 26.52 million GBP (2022)\n- Children: Cristiano Ronaldo Jr., Alana Martina dos Santos Aveiro, Eva Maria Dos Santos, Mateo Ronaldo\n\n!CR7", "# Content\n- URL file containing Goal_no, Season, Competition, Matchday, Venue, Team, Opponent, Result, Position, Minute, At_score, Type_of_goal", "# Featured Notebook\nCR7 - Extensive EDA & Analytics-Cristiano Ronaldo", "# GitHub Project\n- Data Collection : GitHub", "# Download\nkaggle API Command\n\n'!kaggle datasets download -d azminetoushikwasi/cr7-cristiano-ronaldo-all-club-goals-stats'", "## Disclaimer\nThe data collected are all publicly available and it's intended for educational purposes only.", "## Acknowledgement\nCover image credit - URL" ]

[ "license:cc-by-4.0", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-cc-by-4.0 #region-us \n" ]

[ "task_categories:other", "annotations_creators:crowdsourced", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:other", "arxiv:1706.04261", "region:us" ]

[ "1706.04261" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThe Something-Something dataset (version 2) is a collection of 220,847 labeled video clips of humans performing pre-defined, basic actions with everyday objects. It is designed to train machine learning models in fine-grained understanding of human hand gestures like putting something into something, turning something upside down and covering something with something.", "### Supported Tasks and Leaderboards\n\n\n* 'action-recognition': The goal of this task is to classify actions happening in a video. This is a multilabel classification. The leaderboard is available here", "### Languages\n\n\nThe annotations in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'video\\_id': 'str' Unique identifier for each video.\n* 'video': 'str' File object\n* 'placeholders': 'List[str]' Objects present in the video\n* 'text': 'str' Description of what is happening in the video\n* 'labels': 'int' Action found in the video. Indices from 0 to 173.\n\n\n\n\n Click here to see the full list of Something-Something-v2 class labels mapping:\n \n |0 | Approaching something with your camera |\n |1 | Attaching something to something |\n |2 | Bending something so that it deforms |\n |3 | Bending something until it breaks |\n |4 | Burying something in something |\n |5 | Closing something |\n |6 | Covering something with something |\n |7 | Digging something out of something |\n |8 | Dropping something behind something |\n |9 | Dropping something in front of something |\n |10 | Dropping something into something |\n |11 | Dropping something next to something |\n |12 | Dropping something onto something |\n |13 | Failing to put something into something because something does not fit |\n |14 | Folding something |\n |15 | Hitting something with something |\n |16 | Holding something |\n |17 | Holding something behind something |\n |18 | Holding something in front of something |\n |19 | Holding something next to something |\n |20 | Holding something over something |\n |21 | Laying something on the table on its side, not upright |\n |22 | Letting something roll along a flat surface |\n |23 | Letting something roll down a slanted surface |\n |24 | Letting something roll up a slanted surface, so it rolls back down |\n |25 | Lifting a surface with something on it but not enough for it to slide down |\n |26 | Lifting a surface with something on it until it starts sliding down |\n |27 | Lifting something up completely without letting it drop down |\n |28 | Lifting something up completely, then letting it drop down |\n |29 | Lifting something with something on it |\n |30 | Lifting up one end of something without letting it drop down |\n |31 | Lifting up one end of something, then letting it drop down |\n |32 | Moving away from something with your camera |\n |33 | Moving part of something |\n |34 | Moving something across a surface until it falls down |\n |35 | Moving something across a surface without it falling down |\n |36 | Moving something and something away from each other |\n |37 | Moving something and something closer to each other |\n |38 | Moving something and something so they collide with each other |\n |39 | Moving something and something so they pass each other |\n |40 | Moving something away from something |\n |41 | Moving something away from the camera |\n |42 | Moving something closer to something |\n |43 | Moving something down |\n |44 | Moving something towards the camera |\n |45 | Moving something up |\n |46 | Opening something |\n |47 | Picking something up |\n |48 | Piling something up |\n |49 | Plugging something into something |\n |50 | Plugging something into something but pulling it right out as you remove your hand |\n |51 | Poking a hole into some substance |\n |52 | Poking a hole into something soft |\n |53 | Poking a stack of something so the stack collapses |\n |54 | Poking a stack of something without the stack collapsing |\n |55 | Poking something so it slightly moves |\n |56 | Poking something so lightly that it doesn't or almost doesn't move |\n |57 | Poking something so that it falls over |\n |58 | Poking something so that it spins around |\n |59 | Pouring something into something |\n |60 | Pouring something into something until it overflows |\n |61 | Pouring something onto something |\n |62 | Pouring something out of something |\n |63 | Pretending or failing to wipe something off of something |\n |64 | Pretending or trying and failing to twist something |\n |65 | Pretending to be tearing something that is not tearable |\n |66 | Pretending to close something without actually closing it |\n |67 | Pretending to open something without actually opening it |\n |68 | Pretending to pick something up |\n |69 | Pretending to poke something |\n |70 | Pretending to pour something out of something, but something is empty |\n |71 | Pretending to put something behind something |\n |72 | Pretending to put something into something |\n |73 | Pretending to put something next to something |\n |74 | Pretending to put something on a surface |\n |75 | Pretending to put something onto something |\n |76 | Pretending to put something underneath something |\n |77 | Pretending to scoop something up with something |\n |78 | Pretending to spread air onto something |\n |79 | Pretending to sprinkle air onto something |\n |80 | Pretending to squeeze something |\n |81 | Pretending to take something from somewhere |\n |82 | Pretending to take something out of something |\n |83 | Pretending to throw something |\n |84 | Pretending to turn something upside down |\n |85 | Pulling something from behind of something |\n |86 | Pulling something from left to right |\n |87 | Pulling something from right to left |\n |88 | Pulling something onto something |\n |89 | Pulling something out of something |\n |90 | Pulling two ends of something but nothing happens |\n |91 | Pulling two ends of something so that it gets stretched |\n |92 | Pulling two ends of something so that it separates into two pieces |\n |93 | Pushing something from left to right |\n |94 | Pushing something from right to left |\n |95 | Pushing something off of something |\n |96 | Pushing something onto something |\n |97 | Pushing something so it spins |\n |98 | Pushing something so that it almost falls off but doesn't |\n |99 | Pushing something so that it falls off the table |\n |100 | Pushing something so that it slightly moves |\n |101 | Pushing something with something |\n |102 | Putting number of something onto something |\n |103 | Putting something and something on the table |\n |104 | Putting something behind something |\n |105 | Putting something in front of something |\n |106 | Putting something into something |\n |107 | Putting something next to something |\n |108 | Putting something on a flat surface without letting it roll |\n |109 | Putting something on a surface |\n |110 | Putting something on the edge of something so it is not supported and falls down |\n |111 | Putting something onto a slanted surface but it doesn't glide down |\n |112 | Putting something onto something |\n |113 | Putting something onto something else that cannot support it so it falls down |\n |114 | Putting something similar to other things that are already on the table |\n |115 | Putting something that can't roll onto a slanted surface, so it slides down |\n |116 | Putting something that can't roll onto a slanted surface, so it stays where it is |\n |117 | Putting something that cannot actually stand upright upright on the table, so it falls on its side |\n |118 | Putting something underneath something |\n |119 | Putting something upright on the table |\n |120 | Putting something, something and something on the table |\n |121 | Removing something, revealing something behind |\n |122 | Rolling something on a flat surface |\n |123 | Scooping something up with something |\n |124 | Showing a photo of something to the camera |\n |125 | Showing something behind something |\n |126 | Showing something next to something |\n |127 | Showing something on top of something |\n |128 | Showing something to the camera |\n |129 | Showing that something is empty |\n |130 | Showing that something is inside something |\n |131 | Something being deflected from something |\n |132 | Something colliding with something and both are being deflected |\n |133 | Something colliding with something and both come to a halt |\n |134 | Something falling like a feather or paper |\n |135 | Something falling like a rock |\n |136 | Spilling something behind something |\n |137 | Spilling something next to something |\n |138 | Spilling something onto something |\n |139 | Spinning something so it continues spinning |\n |140 | Spinning something that quickly stops spinning |\n |141 | Spreading something onto something |\n |142 | Sprinkling something onto something |\n |143 | Squeezing something |\n |144 | Stacking number of something |\n |145 | Stuffing something into something |\n |146 | Taking one of many similar things on the table |\n |147 | Taking something from somewhere |\n |148 | Taking something out of something |\n |149 | Tearing something into two pieces |\n |150 | Tearing something just a little bit |\n |151 | Throwing something |\n |152 | Throwing something against something |\n |153 | Throwing something in the air and catching it |\n |154 | Throwing something in the air and letting it fall |\n |155 | Throwing something onto a surface |\n |156 | Tilting something with something on it slightly so it doesn't fall down |\n |157 | Tilting something with something on it until it falls off |\n |158 | Tipping something over |\n |159 | Tipping something with something in it over, so something in it falls out |\n |160 | Touching (without moving) part of something |\n |161 | Trying but failing to attach something to something because it doesn't stick |\n |162 | Trying to bend something unbendable so nothing happens |\n |163 | Trying to pour something into something, but missing so it spills next to it |\n |164 | Turning something upside down |\n |165 | Turning the camera downwards while filming something |\n |166 | Turning the camera left while filming something |\n |167 | Turning the camera right while filming something |\n |168 | Turning the camera upwards while filming something |\n |169 | Twisting (wringing) something wet until water comes out |\n |170 | Twisting something |\n |171 | Uncovering something |\n |172 | Unfolding something |\n |173 | Wiping something off of something |", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nFrom the paper:\n\n\n\n> \n> Neural networks trained on datasets such as ImageNet have led to major advances\n> in visual object classification. One obstacle that prevents networks from reasoning more\n> deeply about complex scenes and situations, and from integrating visual knowledge with natural language,\n> like humans do, is their lack of common sense knowledge about the physical world.\n> Videos, unlike still images, contain a wealth of detailed information about the physical world.\n> However, most labelled video datasets represent high-level concepts rather than detailed physical aspects\n> about actions and scenes. In this work, we describe our ongoing collection of the\n> “something-something” database of video prediction tasks whose solutions require a common sense\n> understanding of the depicted situation\n> \n> \n>", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nFrom the paper:\n\n\n\n> \n> As outlined is Section 3 videos available online are largely unsuitable for the goal of learning\n> simple (but finegrained) visual concepts. We therefore ask crowd-workers to provide videos\n> given labels instead of the other way around.\n> \n> \n>", "#### Who are the source language producers?\n\n\nThe dataset authors", "### Annotations", "#### Annotation process\n\n\nThe label is given first and then the video is collected by an AMT worker. More fine-grained details on the process are in the Section 4 of the work.", "#### Who are the annotators?\n\n\nAMT workers", "### Personal and Sensitive Information\n\n\nNothing specifically discussed in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe dataset is useful for action recognition pretraining due to diverse set of actions that happen in it.", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nLicense is a one-page document as defined by QualComm. Please read the license document in detail before using this dataset here.", "### Contributions\n\n\nThanks to @apsdehal for adding this dataset." ]

[ "TAGS\n#task_categories-other #annotations_creators-crowdsourced #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-other #arxiv-1706.04261 #region-us \n", "### Dataset Summary\n\n\nThe Something-Something dataset (version 2) is a collection of 220,847 labeled video clips of humans performing pre-defined, basic actions with everyday objects. It is designed to train machine learning models in fine-grained understanding of human hand gestures like putting something into something, turning something upside down and covering something with something.", "### Supported Tasks and Leaderboards\n\n\n* 'action-recognition': The goal of this task is to classify actions happening in a video. This is a multilabel classification. The leaderboard is available here", "### Languages\n\n\nThe annotations in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields\n\n\n* 'video\\_id': 'str' Unique identifier for each video.\n* 'video': 'str' File object\n* 'placeholders': 'List[str]' Objects present in the video\n* 'text': 'str' Description of what is happening in the video\n* 'labels': 'int' Action found in the video. Indices from 0 to 173.\n\n\n\n\n Click here to see the full list of Something-Something-v2 class labels mapping:\n \n |0 | Approaching something with your camera |\n |1 | Attaching something to something |\n |2 | Bending something so that it deforms |\n |3 | Bending something until it breaks |\n |4 | Burying something in something |\n |5 | Closing something |\n |6 | Covering something with something |\n |7 | Digging something out of something |\n |8 | Dropping something behind something |\n |9 | Dropping something in front of something |\n |10 | Dropping something into something |\n |11 | Dropping something next to something |\n |12 | Dropping something onto something |\n |13 | Failing to put something into something because something does not fit |\n |14 | Folding something |\n |15 | Hitting something with something |\n |16 | Holding something |\n |17 | Holding something behind something |\n |18 | Holding something in front of something |\n |19 | Holding something next to something |\n |20 | Holding something over something |\n |21 | Laying something on the table on its side, not upright |\n |22 | Letting something roll along a flat surface |\n |23 | Letting something roll down a slanted surface |\n |24 | Letting something roll up a slanted surface, so it rolls back down |\n |25 | Lifting a surface with something on it but not enough for it to slide down |\n |26 | Lifting a surface with something on it until it starts sliding down |\n |27 | Lifting something up completely without letting it drop down |\n |28 | Lifting something up completely, then letting it drop down |\n |29 | Lifting something with something on it |\n |30 | Lifting up one end of something without letting it drop down |\n |31 | Lifting up one end of something, then letting it drop down |\n |32 | Moving away from something with your camera |\n |33 | Moving part of something |\n |34 | Moving something across a surface until it falls down |\n |35 | Moving something across a surface without it falling down |\n |36 | Moving something and something away from each other |\n |37 | Moving something and something closer to each other |\n |38 | Moving something and something so they collide with each other |\n |39 | Moving something and something so they pass each other |\n |40 | Moving something away from something |\n |41 | Moving something away from the camera |\n |42 | Moving something closer to something |\n |43 | Moving something down |\n |44 | Moving something towards the camera |\n |45 | Moving something up |\n |46 | Opening something |\n |47 | Picking something up |\n |48 | Piling something up |\n |49 | Plugging something into something |\n |50 | Plugging something into something but pulling it right out as you remove your hand |\n |51 | Poking a hole into some substance |\n |52 | Poking a hole into something soft |\n |53 | Poking a stack of something so the stack collapses |\n |54 | Poking a stack of something without the stack collapsing |\n |55 | Poking something so it slightly moves |\n |56 | Poking something so lightly that it doesn't or almost doesn't move |\n |57 | Poking something so that it falls over |\n |58 | Poking something so that it spins around |\n |59 | Pouring something into something |\n |60 | Pouring something into something until it overflows |\n |61 | Pouring something onto something |\n |62 | Pouring something out of something |\n |63 | Pretending or failing to wipe something off of something |\n |64 | Pretending or trying and failing to twist something |\n |65 | Pretending to be tearing something that is not tearable |\n |66 | Pretending to close something without actually closing it |\n |67 | Pretending to open something without actually opening it |\n |68 | Pretending to pick something up |\n |69 | Pretending to poke something |\n |70 | Pretending to pour something out of something, but something is empty |\n |71 | Pretending to put something behind something |\n |72 | Pretending to put something into something |\n |73 | Pretending to put something next to something |\n |74 | Pretending to put something on a surface |\n |75 | Pretending to put something onto something |\n |76 | Pretending to put something underneath something |\n |77 | Pretending to scoop something up with something |\n |78 | Pretending to spread air onto something |\n |79 | Pretending to sprinkle air onto something |\n |80 | Pretending to squeeze something |\n |81 | Pretending to take something from somewhere |\n |82 | Pretending to take something out of something |\n |83 | Pretending to throw something |\n |84 | Pretending to turn something upside down |\n |85 | Pulling something from behind of something |\n |86 | Pulling something from left to right |\n |87 | Pulling something from right to left |\n |88 | Pulling something onto something |\n |89 | Pulling something out of something |\n |90 | Pulling two ends of something but nothing happens |\n |91 | Pulling two ends of something so that it gets stretched |\n |92 | Pulling two ends of something so that it separates into two pieces |\n |93 | Pushing something from left to right |\n |94 | Pushing something from right to left |\n |95 | Pushing something off of something |\n |96 | Pushing something onto something |\n |97 | Pushing something so it spins |\n |98 | Pushing something so that it almost falls off but doesn't |\n |99 | Pushing something so that it falls off the table |\n |100 | Pushing something so that it slightly moves |\n |101 | Pushing something with something |\n |102 | Putting number of something onto something |\n |103 | Putting something and something on the table |\n |104 | Putting something behind something |\n |105 | Putting something in front of something |\n |106 | Putting something into something |\n |107 | Putting something next to something |\n |108 | Putting something on a flat surface without letting it roll |\n |109 | Putting something on a surface |\n |110 | Putting something on the edge of something so it is not supported and falls down |\n |111 | Putting something onto a slanted surface but it doesn't glide down |\n |112 | Putting something onto something |\n |113 | Putting something onto something else that cannot support it so it falls down |\n |114 | Putting something similar to other things that are already on the table |\n |115 | Putting something that can't roll onto a slanted surface, so it slides down |\n |116 | Putting something that can't roll onto a slanted surface, so it stays where it is |\n |117 | Putting something that cannot actually stand upright upright on the table, so it falls on its side |\n |118 | Putting something underneath something |\n |119 | Putting something upright on the table |\n |120 | Putting something, something and something on the table |\n |121 | Removing something, revealing something behind |\n |122 | Rolling something on a flat surface |\n |123 | Scooping something up with something |\n |124 | Showing a photo of something to the camera |\n |125 | Showing something behind something |\n |126 | Showing something next to something |\n |127 | Showing something on top of something |\n |128 | Showing something to the camera |\n |129 | Showing that something is empty |\n |130 | Showing that something is inside something |\n |131 | Something being deflected from something |\n |132 | Something colliding with something and both are being deflected |\n |133 | Something colliding with something and both come to a halt |\n |134 | Something falling like a feather or paper |\n |135 | Something falling like a rock |\n |136 | Spilling something behind something |\n |137 | Spilling something next to something |\n |138 | Spilling something onto something |\n |139 | Spinning something so it continues spinning |\n |140 | Spinning something that quickly stops spinning |\n |141 | Spreading something onto something |\n |142 | Sprinkling something onto something |\n |143 | Squeezing something |\n |144 | Stacking number of something |\n |145 | Stuffing something into something |\n |146 | Taking one of many similar things on the table |\n |147 | Taking something from somewhere |\n |148 | Taking something out of something |\n |149 | Tearing something into two pieces |\n |150 | Tearing something just a little bit |\n |151 | Throwing something |\n |152 | Throwing something against something |\n |153 | Throwing something in the air and catching it |\n |154 | Throwing something in the air and letting it fall |\n |155 | Throwing something onto a surface |\n |156 | Tilting something with something on it slightly so it doesn't fall down |\n |157 | Tilting something with something on it until it falls off |\n |158 | Tipping something over |\n |159 | Tipping something with something in it over, so something in it falls out |\n |160 | Touching (without moving) part of something |\n |161 | Trying but failing to attach something to something because it doesn't stick |\n |162 | Trying to bend something unbendable so nothing happens |\n |163 | Trying to pour something into something, but missing so it spills next to it |\n |164 | Turning something upside down |\n |165 | Turning the camera downwards while filming something |\n |166 | Turning the camera left while filming something |\n |167 | Turning the camera right while filming something |\n |168 | Turning the camera upwards while filming something |\n |169 | Twisting (wringing) something wet until water comes out |\n |170 | Twisting something |\n |171 | Uncovering something |\n |172 | Unfolding something |\n |173 | Wiping something off of something |", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nFrom the paper:\n\n\n\n> \n> Neural networks trained on datasets such as ImageNet have led to major advances\n> in visual object classification. One obstacle that prevents networks from reasoning more\n> deeply about complex scenes and situations, and from integrating visual knowledge with natural language,\n> like humans do, is their lack of common sense knowledge about the physical world.\n> Videos, unlike still images, contain a wealth of detailed information about the physical world.\n> However, most labelled video datasets represent high-level concepts rather than detailed physical aspects\n> about actions and scenes. In this work, we describe our ongoing collection of the\n> “something-something” database of video prediction tasks whose solutions require a common sense\n> understanding of the depicted situation\n> \n> \n>", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nFrom the paper:\n\n\n\n> \n> As outlined is Section 3 videos available online are largely unsuitable for the goal of learning\n> simple (but finegrained) visual concepts. We therefore ask crowd-workers to provide videos\n> given labels instead of the other way around.\n> \n> \n>", "#### Who are the source language producers?\n\n\nThe dataset authors", "### Annotations", "#### Annotation process\n\n\nThe label is given first and then the video is collected by an AMT worker. More fine-grained details on the process are in the Section 4 of the work.", "#### Who are the annotators?\n\n\nAMT workers", "### Personal and Sensitive Information\n\n\nNothing specifically discussed in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThe dataset is useful for action recognition pretraining due to diverse set of actions that happen in it.", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nLicense is a one-page document as defined by QualComm. Please read the license document in detail before using this dataset here.", "### Contributions\n\n\nThanks to @apsdehal for adding this dataset." ]

[ "region:us" ]

[]

[]

[ "# Toxic Conversation\r\nThis is a version of the Jigsaw Unintended Bias in Toxicity Classification dataset. It contains comments from the Civil Comments platform together with annotations if the comment is toxic or not.\r\n\r\nThis dataset just contains the first 50k training examples.\r\n\r\n10 annotators annotated each example and, as recommended in the task page, set a comment as toxic when target >= 0.5\r\n\r\nThe dataset is inbalanced, with only about 8% of the comments marked as toxic." ]

[ "TAGS\n#region-us \n", "# Toxic Conversation\r\nThis is a version of the Jigsaw Unintended Bias in Toxicity Classification dataset. It contains comments from the Civil Comments platform together with annotations if the comment is toxic or not.\r\n\r\nThis dataset just contains the first 50k training examples.\r\n\r\n10 annotators annotated each example and, as recommended in the task page, set a comment as toxic when target >= 0.5\r\n\r\nThe dataset is inbalanced, with only about 8% of the comments marked as toxic." ]

[ "license:mit", "region:us" ]

[]

[]

[ "# Dataset Card for \"20-Newsgroups\"" ]

[ "TAGS\n#license-mit #region-us \n", "# Dataset Card for \"20-Newsgroups\"" ]

[ "task_ids:closed-domain-qa", "annotations_creators:expert-generated", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:10k<n<100K", "source_datasets:original", "language:en", "license:other", "arxiv:1705.00754", "region:us" ]

[ "1705.00754" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThe ActivityNet Captions dataset connects videos to a series of temporally annotated sentence descriptions. Each sentence covers an unique segment of the video, describing multiple events that occur. These events may occur over very long or short periods of time and are not limited in any capacity, allowing them to co-occur. On average, each of the 20k videos contains 3.65 temporally localized sentences, resulting in a total of 100k sentences. We find that the number of sentences per video follows a relatively normal distribution. Furthermore, as the video duration increases, the number of sentences also increases. Each sentence has an average length of 13.48 words, which is also normally distributed. You can find more details of the dataset under the ActivityNet Captions Dataset section, and under supplementary materials in the paper.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_id' : 'str' unique identifier for the video\n* 'video\\_path': 'str' Path to the video file\n-'duration': 'float32' Duration of the video\n* 'captions\\_starts': 'List\\_float32' List of timestamps denoting the time at which each caption starts\n* 'captions\\_ends': 'List\\_float32' List of timestamps denoting the time at which each caption ends\n* 'en\\_captions': 'list\\_str' List of english captions describing parts of the video", "### Data Splits", "### Annotations\n\n\nQuoting ActivityNet Captions' paper: \n\n\"Each annotation task was divided into two steps: (1)\nWriting a paragraph describing all major events happening\nin the videos in a paragraph, with each sentence of the paragraph describing one event, and (2) Labeling the\nstart and end time in the video in which each sentence in the\nparagraph event occurred.\"", "### Who annotated the dataset?\n\n\nAmazon Mechnical Turk annotators", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "TAGS\n#task_ids-closed-domain-qa #annotations_creators-expert-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-10k<n<100K #source_datasets-original #language-English #license-other #arxiv-1705.00754 #region-us \n", "### Dataset Summary\n\n\nThe ActivityNet Captions dataset connects videos to a series of temporally annotated sentence descriptions. Each sentence covers an unique segment of the video, describing multiple events that occur. These events may occur over very long or short periods of time and are not limited in any capacity, allowing them to co-occur. On average, each of the 20k videos contains 3.65 temporally localized sentences, resulting in a total of 100k sentences. We find that the number of sentences per video follows a relatively normal distribution. Furthermore, as the video duration increases, the number of sentences also increases. Each sentence has an average length of 13.48 words, which is also normally distributed. You can find more details of the dataset under the ActivityNet Captions Dataset section, and under supplementary materials in the paper.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_id' : 'str' unique identifier for the video\n* 'video\\_path': 'str' Path to the video file\n-'duration': 'float32' Duration of the video\n* 'captions\\_starts': 'List\\_float32' List of timestamps denoting the time at which each caption starts\n* 'captions\\_ends': 'List\\_float32' List of timestamps denoting the time at which each caption ends\n* 'en\\_captions': 'list\\_str' List of english captions describing parts of the video", "### Data Splits", "### Annotations\n\n\nQuoting ActivityNet Captions' paper: \n\n\"Each annotation task was divided into two steps: (1)\nWriting a paragraph describing all major events happening\nin the videos in a paragraph, with each sentence of the paragraph describing one event, and (2) Labeling the\nstart and end time in the video in which each sentence in the\nparagraph event occurred.\"", "### Who annotated the dataset?\n\n\nAmazon Mechnical Turk annotators", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "language:en", "region:us" ]

[]

[ "en" ]

[ "### Languages\n\n\nThe BCP-47 code for the dataset's language is en.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nA sample from this dataset looks as follows:", "### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):", "### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ "TAGS\n#language-English #region-us \n", "### Languages\n\n\nThe BCP-47 code for the dataset's language is en.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nA sample from this dataset looks as follows:", "### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):", "### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ "license:zlib", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-zlib #region-us \n" ]

[ "task_categories:text-classification", "task_ids:topic-classification", "task_ids:semantic-similarity-classification", "language_creators:found", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:cc-by-4.0", "Cognitive Control", "PubMed", "arxiv:2203.11016", "doi:10.57967/hf/0548", "region:us" ]

[ "2203.11016" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThe 2021 dataset, collected in December 2021, contains 385,705 distinct scientific articles, featuring their title, abstract, relevant metadata, and embeddings.\nThe articles were specifically selected for their relevance to cognitive control constructs and associated tasks.", "### Supported Tasks and Leaderboards\n\n\nTopic Modeling, Text Embedding", "### Languages\n\n\nEnglish\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n522,972 scientific articles, of which 385,705 are unique.", "### Data Fields\n\n\nThe CSV files contain the following fields:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "### Annotations", "#### Annotation process", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information", "### Acknowledgments\n\n\nThis research was supported by the Luxembourg National Research Fund (ATTRACT/2016/ID/11242114/DIGILEARN and INTER Mobility/2017-2/ID/11765868/ULALA).\n\n\nTo cite the paper use the following entry:" ]

[ "TAGS\n#task_categories-text-classification #task_ids-topic-classification #task_ids-semantic-similarity-classification #language_creators-found #language_creators-expert-generated #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-cc-by-4.0 #Cognitive Control #PubMed #arxiv-2203.11016 #doi-10.57967/hf/0548 #region-us \n", "### Dataset Summary\n\n\nThe 2021 dataset, collected in December 2021, contains 385,705 distinct scientific articles, featuring their title, abstract, relevant metadata, and embeddings.\nThe articles were specifically selected for their relevance to cognitive control constructs and associated tasks.", "### Supported Tasks and Leaderboards\n\n\nTopic Modeling, Text Embedding", "### Languages\n\n\nEnglish\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n522,972 scientific articles, of which 385,705 are unique.", "### Data Fields\n\n\nThe CSV files contain the following fields:", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "### Annotations", "#### Annotation process", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information", "### Acknowledgments\n\n\nThis research was supported by the Luxembourg National Research Fund (ATTRACT/2016/ID/11242114/DIGILEARN and INTER Mobility/2017-2/ID/11765868/ULALA).\n\n\nTo cite the paper use the following entry:" ]

[ "size_categories:100K<n<1M", "language:en", "music", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#size_categories-100K<n<1M #language-English #music #region-us \n" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ "task_categories:text2text-generation", "task_categories:text-generation", "task_ids:language-modeling", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "language:id", "license:mit", "region:us" ]

[]

[ "id" ]

[ "### Dataset Summary\n\n\nIndonesian foods are well-known for their rich taste. There are many spices used even for daily foods. This dataset may give insight on how to prepare Indonesian food.\n\n\nid\\_recipe is an Indonesian Food Recipe dataset. The dataset contains >10000 Indonesian Recipe.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nIndonesian", "### Data Splits\n\n\nHere are the number of examples", "### Source Data\n\n\nhere", "### Annotations", "#### Annotation process\n\n\n[N/A]", "#### Who are the annotators?\n\n\n[N/A]", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nMIT License\n\n\n[N/A]", "### Contributions\n\n\nThanks to @sultan for adding this dataset" ]

[ "TAGS\n#task_categories-text2text-generation #task_categories-text-generation #task_ids-language-modeling #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-Indonesian #license-mit #region-us \n", "### Dataset Summary\n\n\nIndonesian foods are well-known for their rich taste. There are many spices used even for daily foods. This dataset may give insight on how to prepare Indonesian food.\n\n\nid\\_recipe is an Indonesian Food Recipe dataset. The dataset contains >10000 Indonesian Recipe.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nIndonesian", "### Data Splits\n\n\nHere are the number of examples", "### Source Data\n\n\nhere", "### Annotations", "#### Annotation process\n\n\n[N/A]", "#### Who are the annotators?\n\n\n[N/A]", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nMIT License\n\n\n[N/A]", "### Contributions\n\n\nThanks to @sultan for adding this dataset" ]

[ "task_categories:text-classification", "task_categories:text-generation", "task_ids:multi-label-classification", "task_ids:multi-input-text-classification", "task_ids:language-modeling", "annotations_creators:other", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for Slither Audited Smart Contracts", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Point of Contact: Martina Rossini", "### Dataset Summary\n\nThis dataset contains source code and deployed bytecode for Solidity Smart Contracts that have been verified on URL, along with a classification of their vulnerabilities according to the Slither static analysis framework.", "### Supported Tasks and Leaderboards\n\n- 'text-classification': The dataset can be used to train a model for both binary and multilabel text classification on smart contracts bytecode and source code. The model performance is evaluated based on the accuracy of the predicted labels as compared to the given labels in the dataset.\n- 'text-generation': The dataset can also be used to train a language model for the Solidity programming language\n- 'image-classification': By pre-processing the bytecode data to obtain RGB images, the dataset can also be used to train convolutional neural networks for code vulnerability detection and classification.", "### Languages\n\nThe language annotations are in English, while all the source codes are in Solidity.", "## Dataset Structure", "### Data Instances\n\nEach data instance contains the following features: 'address', 'source_code' and 'bytecode'. The label comes in two configuration, either a plain-text cleaned up version of the output given by the Slither tool or a multi-label version, which consists in a simple list of integers, each one representing a particular vulnerability class. Label 4 indicates that the contract is safe.\n\nAn example from a plain-text configuration looks as follows:\n\n\nAn example from a multi-label configuration looks as follows:", "### Data Fields\n\n- 'address': a string representing the address of the smart contract deployed on the Ethereum main net\n- 'source_code': a flattened version of the smart contract codebase in Solidity\n- 'bytecode': a string representing the smart contract's bytecode, obtained when calling 'URL.getCode()'. Note that in some cases where this was not available, the string is simply '0x'.\n- 'slither': either a cleaned up version of Slither's JSON output or a list of class labels", "### Data Splits\n\nThe dataset comes in 6 configurations and train, test and validation splits are only provided for those configurations that do not include 'all-' in their names. Test and Validation splits are both about 15% of the total.", "## Dataset Creation", "### Curation Rationale\n\nslither-audited-smart-contracts was built to provide a freely available large scale dataset for vulnerability detection and classification on verified Solidity smart contracts. Indeed, the biggest open source dataset for this task at the moment of writing is SmartBugs Wild, containing 47,398 smart contracts that were labeled with 9 tools withing the SmartBugs framework.", "### Source Data", "#### Initial Data Collection and Normalization\n\nThe dataset was constructed started from the list of verified smart contracts provided at Smart Contract Sanctuary. Then, smart contract source code was either downloaded from the aforementioned repo or downloaded via Etherscan and flattened using the Slither contract flattener. The bytecode was downloaded using the URL library, in particular the 'URL.getCode()' function and using INFURA as our endpoint.\nFinally, every smart contract was analyzed using the Slither static analysis framework. The tool found 38 different vulnerability classes in the collected contracts and they were then mapped to 9 labels according to what is shown in the file 'label_mappings.json'. These mappings were derived by following the guidelines at Decentralized Application Security Project (DASP) and at Smart Contract Weakness Classification Registry. They were also inspired by the mappings used for Slither's detection by the team that labeled the SmartBugs Wild dataset, which can be found here.", "## Additional Information", "### Dataset Curators\n\nThe dataset was initially created by Martina Rossini during work done for the project of the course Blockchain and Cryptocurrencies of the University of Bologna (Italy).", "### Licensing Information\n\nThe license in the file LICENSE applies to all the files in this repository, except for the Solidity source code of the contracts. These are still publicly available, were obtained using the Etherscan APIs, and retain their original licenses.\n\n\n\nIf you are using this dataset in your research and paper, here's how you can cite it:", "### Contributions\nThanks to @mwritescode for adding this dataset." ]

[ "TAGS\n#task_categories-text-classification #task_categories-text-generation #task_ids-multi-label-classification #task_ids-multi-input-text-classification #task_ids-language-modeling #annotations_creators-other #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-mit #region-us \n", "# Dataset Card for Slither Audited Smart Contracts", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Point of Contact: Martina Rossini", "### Dataset Summary\n\nThis dataset contains source code and deployed bytecode for Solidity Smart Contracts that have been verified on URL, along with a classification of their vulnerabilities according to the Slither static analysis framework.", "### Supported Tasks and Leaderboards\n\n- 'text-classification': The dataset can be used to train a model for both binary and multilabel text classification on smart contracts bytecode and source code. The model performance is evaluated based on the accuracy of the predicted labels as compared to the given labels in the dataset.\n- 'text-generation': The dataset can also be used to train a language model for the Solidity programming language\n- 'image-classification': By pre-processing the bytecode data to obtain RGB images, the dataset can also be used to train convolutional neural networks for code vulnerability detection and classification.", "### Languages\n\nThe language annotations are in English, while all the source codes are in Solidity.", "## Dataset Structure", "### Data Instances\n\nEach data instance contains the following features: 'address', 'source_code' and 'bytecode'. The label comes in two configuration, either a plain-text cleaned up version of the output given by the Slither tool or a multi-label version, which consists in a simple list of integers, each one representing a particular vulnerability class. Label 4 indicates that the contract is safe.\n\nAn example from a plain-text configuration looks as follows:\n\n\nAn example from a multi-label configuration looks as follows:", "### Data Fields\n\n- 'address': a string representing the address of the smart contract deployed on the Ethereum main net\n- 'source_code': a flattened version of the smart contract codebase in Solidity\n- 'bytecode': a string representing the smart contract's bytecode, obtained when calling 'URL.getCode()'. Note that in some cases where this was not available, the string is simply '0x'.\n- 'slither': either a cleaned up version of Slither's JSON output or a list of class labels", "### Data Splits\n\nThe dataset comes in 6 configurations and train, test and validation splits are only provided for those configurations that do not include 'all-' in their names. Test and Validation splits are both about 15% of the total.", "## Dataset Creation", "### Curation Rationale\n\nslither-audited-smart-contracts was built to provide a freely available large scale dataset for vulnerability detection and classification on verified Solidity smart contracts. Indeed, the biggest open source dataset for this task at the moment of writing is SmartBugs Wild, containing 47,398 smart contracts that were labeled with 9 tools withing the SmartBugs framework.", "### Source Data", "#### Initial Data Collection and Normalization\n\nThe dataset was constructed started from the list of verified smart contracts provided at Smart Contract Sanctuary. Then, smart contract source code was either downloaded from the aforementioned repo or downloaded via Etherscan and flattened using the Slither contract flattener. The bytecode was downloaded using the URL library, in particular the 'URL.getCode()' function and using INFURA as our endpoint.\nFinally, every smart contract was analyzed using the Slither static analysis framework. The tool found 38 different vulnerability classes in the collected contracts and they were then mapped to 9 labels according to what is shown in the file 'label_mappings.json'. These mappings were derived by following the guidelines at Decentralized Application Security Project (DASP) and at Smart Contract Weakness Classification Registry. They were also inspired by the mappings used for Slither's detection by the team that labeled the SmartBugs Wild dataset, which can be found here.", "## Additional Information", "### Dataset Curators\n\nThe dataset was initially created by Martina Rossini during work done for the project of the course Blockchain and Cryptocurrencies of the University of Bologna (Italy).", "### Licensing Information\n\nThe license in the file LICENSE applies to all the files in this repository, except for the Solidity source code of the contracts. These are still publicly available, were obtained using the Etherscan APIs, and retain their original licenses.\n\n\n\nIf you are using this dataset in your research and paper, here's how you can cite it:", "### Contributions\nThanks to @mwritescode for adding this dataset." ]

[ "task_categories:text-classification", "annotations_creators:weak supervision", "annotations_creators:expert-generated", "multilinguality:monolingual", "size_categories:1K<n<10K", "size_categories:10K<n<100K", "source_datasets:original", "language:en", "license:unknown", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for [products-2017]", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Annotations\n- Additional Information\n - Citation Information", "## Dataset Description\n\n- Homepage: LSPCv2 Homepage\n- Point of Contact: Ralph Peeters", "### Dataset Summary\n\nMany e-shops have started to mark-up product data within their HTML pages using the URL vocabulary. The Web Data Commons project regularly extracts such data from the Common Crawl, a large public web crawl. The Web Data Commons Training and Test Sets for Large-Scale Product Matching contain product offers from different e-shops in the form of binary product pairs (with corresponding label \"match\" or \"no match\")\n\nIn order to support the evaluation of machine learning-based matching methods, the data is split into training, validation and test set. We provide training and validation sets in four different sizes for four product categories. The labels of the test sets were manually checked while those of the training sets were derived using shared product identifiers from the Web via weak supervision.\n\nThe data stems from the WDC Product Data Corpus for Large-Scale Product Matching - Version 2.0 which consists of 26 million product offers originating from 79 thousand websites.", "### Supported Tasks and Leaderboards\n\nEntity Matching, Product Matching", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nThe data is structured as pairs of product offers with the corresponding match/non-match label. This is an example instance from the computers category:", "### Data Fields\n\n- pair_id: unique identifier of a pair (string)\n- label: binary label, match or non-match (int)\n\nThe following attributes are contained twice, once for the first and once for the second product offer\n\n- id: unique id of the product offer (int)\n- category: product category (string)\n- cluster_id: id of the product cluster from the original corpus this offer belongs to (int)\n- brand: brand of the product (string)\n- title: product title (string)\n- description: longer product description (string)\n- price: price of the product offer (string)\n- specTableContent: additional data found in specification tables on the webpage that contains the product offer (string)", "### Data Splits\n- Computers\n - Test set - 1100 pairs\n - Small Train set - 2267 pairs\n - Small Validation set - 567 pairs\n - Medium Train set - 6475 pairs\n - Medium Validation set - 1619 pairs\n - Large Train set - 26687 pairs\n - Large Validation set - 6672 pairs\n - XLarge Train set - 54768 pairs\n - Xlarge Validation set - 13693 pairs\n\n- Cameras\n - Test set - 1100 pairs\n - Small Train set - 1508 pairs\n - Small Validation set - 378 pairs\n - Medium Train set - 4204 pairs\n - Medium Validation set - 1051 pairs\n - Large Train set - 16028 pairs\n - Large Validation set - 4008 pairs\n - XLarge Train set - 33821 pairs\n - Xlarge Validation set - 8456 pairs\n\n- Watches\n - Test set - 1100 pairs\n - Small Train set - 1804 pairs\n - Small Validation set - 451 pairs\n - Medium Train set - 5130 pairs\n - Medium Validation set - 1283 pairs\n - Large Train set - 21621 pairs\n - Large Validation set - 5406 pairs\n - XLarge Train set - 49255 pairs\n - Xlarge Validation set - 12314 pairs\n\n- Shoes\n - Test set - 1100 pairs\n - Small Train set - 1650 pairs\n - Small Validation set - 413 pairs\n - Medium Train set - 4644 pairs\n - Medium Validation set - 1161 pairs\n - Large Train set - 18391 pairs\n - Large Validation set - 4598 pairs\n - XLarge Train set - 33943 pairs\n - Xlarge Validation set - 8486 pairs", "## Dataset Creation", "### Annotations", "#### Annotation process\n\n- Training and Validation sets: distant supervision via shared URL product IDs\n- Test sets: Single expert annotator", "#### Who are the annotators?\n\nRalph Peeters", "## Additional Information" ]

[ "TAGS\n#task_categories-text-classification #annotations_creators-weak supervision #annotations_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #size_categories-10K<n<100K #source_datasets-original #language-English #license-unknown #region-us \n", "# Dataset Card for [products-2017]", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Annotations\n- Additional Information\n - Citation Information", "## Dataset Description\n\n- Homepage: LSPCv2 Homepage\n- Point of Contact: Ralph Peeters", "### Dataset Summary\n\nMany e-shops have started to mark-up product data within their HTML pages using the URL vocabulary. The Web Data Commons project regularly extracts such data from the Common Crawl, a large public web crawl. The Web Data Commons Training and Test Sets for Large-Scale Product Matching contain product offers from different e-shops in the form of binary product pairs (with corresponding label \"match\" or \"no match\")\n\nIn order to support the evaluation of machine learning-based matching methods, the data is split into training, validation and test set. We provide training and validation sets in four different sizes for four product categories. The labels of the test sets were manually checked while those of the training sets were derived using shared product identifiers from the Web via weak supervision.\n\nThe data stems from the WDC Product Data Corpus for Large-Scale Product Matching - Version 2.0 which consists of 26 million product offers originating from 79 thousand websites.", "### Supported Tasks and Leaderboards\n\nEntity Matching, Product Matching", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nThe data is structured as pairs of product offers with the corresponding match/non-match label. This is an example instance from the computers category:", "### Data Fields\n\n- pair_id: unique identifier of a pair (string)\n- label: binary label, match or non-match (int)\n\nThe following attributes are contained twice, once for the first and once for the second product offer\n\n- id: unique id of the product offer (int)\n- category: product category (string)\n- cluster_id: id of the product cluster from the original corpus this offer belongs to (int)\n- brand: brand of the product (string)\n- title: product title (string)\n- description: longer product description (string)\n- price: price of the product offer (string)\n- specTableContent: additional data found in specification tables on the webpage that contains the product offer (string)", "### Data Splits\n- Computers\n - Test set - 1100 pairs\n - Small Train set - 2267 pairs\n - Small Validation set - 567 pairs\n - Medium Train set - 6475 pairs\n - Medium Validation set - 1619 pairs\n - Large Train set - 26687 pairs\n - Large Validation set - 6672 pairs\n - XLarge Train set - 54768 pairs\n - Xlarge Validation set - 13693 pairs\n\n- Cameras\n - Test set - 1100 pairs\n - Small Train set - 1508 pairs\n - Small Validation set - 378 pairs\n - Medium Train set - 4204 pairs\n - Medium Validation set - 1051 pairs\n - Large Train set - 16028 pairs\n - Large Validation set - 4008 pairs\n - XLarge Train set - 33821 pairs\n - Xlarge Validation set - 8456 pairs\n\n- Watches\n - Test set - 1100 pairs\n - Small Train set - 1804 pairs\n - Small Validation set - 451 pairs\n - Medium Train set - 5130 pairs\n - Medium Validation set - 1283 pairs\n - Large Train set - 21621 pairs\n - Large Validation set - 5406 pairs\n - XLarge Train set - 49255 pairs\n - Xlarge Validation set - 12314 pairs\n\n- Shoes\n - Test set - 1100 pairs\n - Small Train set - 1650 pairs\n - Small Validation set - 413 pairs\n - Medium Train set - 4644 pairs\n - Medium Validation set - 1161 pairs\n - Large Train set - 18391 pairs\n - Large Validation set - 4598 pairs\n - XLarge Train set - 33943 pairs\n - Xlarge Validation set - 8486 pairs", "## Dataset Creation", "### Annotations", "#### Annotation process\n\n- Training and Validation sets: distant supervision via shared URL product IDs\n- Test sets: Single expert annotator", "#### Who are the annotators?\n\nRalph Peeters", "## Additional Information" ]

[ "arxiv:2005.06557", "region:us" ]

[ "2005.06557" ]

[]

[ "### Dataset Summary\n\n\nWe present QADI, an automatically collected dataset of tweets belonging to a wide range of\ncountry-level Arabic dialects \u0014covering 18 different countries in the Middle East and North\nAfrica region. Our method for building this dataset relies on applying multiple filters to identify\nusers who belong to different countries based on their account descriptions and to eliminate\ntweets that are either written in Modern Standard Arabic or contain inappropriate language. The\nresultant dataset contains 540k tweets from 2,525 users who are evenly distributed across 18 Arab countries.", "### Supported Tasks and Leaderboards\n\n\n* Multi-class-Classification: Using extrinsic evaluation, we are able to build effective country-level dialect identification on tweets with a macro-averaged F1-score of 51.5% across 18 classes.\nArabic-Dialect-Identification, rather than what used in the paper Using intrinsic evaluation, they show that the labels of a set of randomly selected tweets are 91.5% accurate. For extrinsic evaluation, they are able to build effective country-level dialect identification on tweets with a macro-averaged F1-score of 60.6% across 18 classes Paper. And we aimed by next work to fine tune models with that data to see how the result will be.", "### Languages\n\n\nArabic\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n'{\"id\": [1159906099585327104, 950123809608171648, 1091295506960142336], \"label\": [10, 14, 2], \"text\": [\"ايه الخيبة و الهرتلة قدام الجون دول؟؟ \\U0001f92a\\nالعيال دي تتعلق في الفلكة يا معلم كلوب\", \"@FIA\\_WIS تذكرت ما اسمي عائشة انا اسمي خولة\", \"@showqiy @3nood\\_mh لا والله نروح نشجع قطر و نفرح معهم وش رايك بعد\"]}'", "### Data Fields\n\n\n'\"{'id': Value(dtype='int64', id=None), 'label': ClassLabel(num\\_classes=18, names=['OM', 'SD', 'SA', 'KW', 'QA', 'LB', 'JO', 'SY', 'IQ', 'MA', 'EG', 'PL', 'YE', 'BH', 'DZ', 'AE', 'TN', 'LY'], id=None), 'text': Value(dtype='string', id=None)}\"'", "### Data Splits\n\n\nThis dataset is split into a train, validation and test split. The split sizes are as follow:\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\n{aabdelali,hmubarak,ysamih,sahassan2,kdarwish}@URL", "### Licensing Information\n\n\n@unknown{unknown,\nauthor = {Abdelali, Ahmed and Mubarak, Hamdy and Samih, Younes and Hassan, Sabit and Darwish, Kareem},\nyear = {2020},\nmonth = {05},\npages = {},\ntitle = {Arabic Dialect Identification in the Wild}\n}" ]

[ "TAGS\n#arxiv-2005.06557 #region-us \n", "### Dataset Summary\n\n\nWe present QADI, an automatically collected dataset of tweets belonging to a wide range of\ncountry-level Arabic dialects \u0014covering 18 different countries in the Middle East and North\nAfrica region. Our method for building this dataset relies on applying multiple filters to identify\nusers who belong to different countries based on their account descriptions and to eliminate\ntweets that are either written in Modern Standard Arabic or contain inappropriate language. The\nresultant dataset contains 540k tweets from 2,525 users who are evenly distributed across 18 Arab countries.", "### Supported Tasks and Leaderboards\n\n\n* Multi-class-Classification: Using extrinsic evaluation, we are able to build effective country-level dialect identification on tweets with a macro-averaged F1-score of 51.5% across 18 classes.\nArabic-Dialect-Identification, rather than what used in the paper Using intrinsic evaluation, they show that the labels of a set of randomly selected tweets are 91.5% accurate. For extrinsic evaluation, they are able to build effective country-level dialect identification on tweets with a macro-averaged F1-score of 60.6% across 18 classes Paper. And we aimed by next work to fine tune models with that data to see how the result will be.", "### Languages\n\n\nArabic\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\n'{\"id\": [1159906099585327104, 950123809608171648, 1091295506960142336], \"label\": [10, 14, 2], \"text\": [\"ايه الخيبة و الهرتلة قدام الجون دول؟؟ \\U0001f92a\\nالعيال دي تتعلق في الفلكة يا معلم كلوب\", \"@FIA\\_WIS تذكرت ما اسمي عائشة انا اسمي خولة\", \"@showqiy @3nood\\_mh لا والله نروح نشجع قطر و نفرح معهم وش رايك بعد\"]}'", "### Data Fields\n\n\n'\"{'id': Value(dtype='int64', id=None), 'label': ClassLabel(num\\_classes=18, names=['OM', 'SD', 'SA', 'KW', 'QA', 'LB', 'JO', 'SY', 'IQ', 'MA', 'EG', 'PL', 'YE', 'BH', 'DZ', 'AE', 'TN', 'LY'], id=None), 'text': Value(dtype='string', id=None)}\"'", "### Data Splits\n\n\nThis dataset is split into a train, validation and test split. The split sizes are as follow:\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\n{aabdelali,hmubarak,ysamih,sahassan2,kdarwish}@URL", "### Licensing Information\n\n\n@unknown{unknown,\nauthor = {Abdelali, Ahmed and Mubarak, Hamdy and Samih, Younes and Hassan, Sabit and Darwish, Kareem},\nyear = {2020},\nmonth = {05},\npages = {},\ntitle = {Arabic Dialect Identification in the Wild}\n}" ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for \"transformers_issues_labels\"\n\nMore Information needed" ]

[ "TAGS\n#region-us \n", "# Dataset Card for \"transformers_issues_labels\"\n\nMore Information needed" ]

[ "arxiv:2106.06147", "region:us" ]

[ "2106.06147" ]

[]

[ "# CLEAR2 dataset\n\nThis dataset was presented in the article \"NAAQA: A Neural Architecture for Acoustic Question answering\" submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence in 2021.\nURL\n\nThe code to generate this dataset is available at : URL", "## Structure\n- scenes/ : 1 json file per set (Train/val/test)\n - Specify the order and the timings of each sounds in a scene\n- questions/ : 1 json files per set (Train/val/test).\n - Specify the questions and answers for each scenes.\n - The functional program of the question is also provided\n- audio/ : Acoustic scenes recordings (FLAC)\n - train/\n - val/\n - test/\n- URL : List all possible answers (Split by question categories)" ]

[ "TAGS\n#arxiv-2106.06147 #region-us \n", "# CLEAR2 dataset\n\nThis dataset was presented in the article \"NAAQA: A Neural Architecture for Acoustic Question answering\" submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence in 2021.\nURL\n\nThe code to generate this dataset is available at : URL", "## Structure\n- scenes/ : 1 json file per set (Train/val/test)\n - Specify the order and the timings of each sounds in a scene\n- questions/ : 1 json files per set (Train/val/test).\n - Specify the questions and answers for each scenes.\n - The functional program of the question is also provided\n- audio/ : Acoustic scenes recordings (FLAC)\n - train/\n - val/\n - test/\n- URL : List all possible answers (Split by question categories)" ]

[ "region:us" ]

[]

[]

[ "### Dataset Summary\n\n\nThis dataset was created based on metadata for mined bitext released by Meta AI. It contains bitext for 248 pairs for the African languages that are part of the 2022 WMT Shared Task on Large Scale Machine Translation Evaluation for African Languages.", "#### How to use the data\n\n\nThere are two ways to access the data:\n\n\n* Via the Hugging Face Python datasets library\n* Clone the git repo", "### Supported Tasks and Leaderboards\n\n\nThis dataset is one of resources allowed under the Constrained Track for the 2022 WMT Shared Task on Large Scale Machine Translation Evaluation for African Languages.", "### Languages", "#### Focus languages\n\n\n\nColonial linguae francae: English - eng, French - fra\n\n\nDataset Structure\n-----------------\n\n\nThe dataset contains gzipped tab delimited text files for each direction. Each text file contains lines with parallel sentences.", "### Data Instances\n\n\nThe dataset contains 248 language pairs.\n\n\nSentence counts for each pair can be found here.", "### Data Fields\n\n\nEvery instance for a language pair contains the following fields: 'translation' (containing sentence pairs), 'laser\\_score', 'source\\_sentence\\_lid', 'target\\_sentence\\_lid', where 'lid' is language classification probability.\n\n\nExample:", "### Data Splits\n\n\nThe data is not split into train, dev, and test.\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nParallel sentences from monolingual data in Common Crawl and ParaCrawl were identified via Language-Agnostic Sentence Representation (LASER) encoders.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nMonolingual data was obtained from Common Crawl and ParaCrawl.", "#### Who are the source language producers?\n\n\nContributors to web text in Common Crawl and ParaCrawl.", "### Annotations", "#### Annotation process\n\n\nThe data was not human annotated. The metadata used to create the dataset can be found here: URL", "#### Who are the annotators?\n\n\nThe data was not human annotated. Parallel text from Common Crawl and Para Crawl monolingual data were identified automatically via LASER encoders.", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThis dataset provides data for training machine learning systems for many languages that have low resources available for NLP.", "### Discussion of Biases\n\n\nBiases in the data have not been studied.", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThe dataset is released under the terms of ODC-BY. By using this, you are also bound by the Internet Archive Terms of Use in respect of the content contained in the dataset.\n\n\nNLLB Team et al, No Language Left Behind: Scaling Human-Centered Machine Translation, Arxiv, 2022.", "### Contributions\n\n\nWe thank the AllenNLP team at AI2 for hosting and releasing this data, including Akshita Bhagia (for engineering efforts to create the huggingface dataset), and Jesse Dodge (for organizing the connection)." ]

[ "TAGS\n#region-us \n", "### Dataset Summary\n\n\nThis dataset was created based on metadata for mined bitext released by Meta AI. It contains bitext for 248 pairs for the African languages that are part of the 2022 WMT Shared Task on Large Scale Machine Translation Evaluation for African Languages.", "#### How to use the data\n\n\nThere are two ways to access the data:\n\n\n* Via the Hugging Face Python datasets library\n* Clone the git repo", "### Supported Tasks and Leaderboards\n\n\nThis dataset is one of resources allowed under the Constrained Track for the 2022 WMT Shared Task on Large Scale Machine Translation Evaluation for African Languages.", "### Languages", "#### Focus languages\n\n\n\nColonial linguae francae: English - eng, French - fra\n\n\nDataset Structure\n-----------------\n\n\nThe dataset contains gzipped tab delimited text files for each direction. Each text file contains lines with parallel sentences.", "### Data Instances\n\n\nThe dataset contains 248 language pairs.\n\n\nSentence counts for each pair can be found here.", "### Data Fields\n\n\nEvery instance for a language pair contains the following fields: 'translation' (containing sentence pairs), 'laser\\_score', 'source\\_sentence\\_lid', 'target\\_sentence\\_lid', where 'lid' is language classification probability.\n\n\nExample:", "### Data Splits\n\n\nThe data is not split into train, dev, and test.\n\n\nDataset Creation\n----------------", "### Curation Rationale\n\n\nParallel sentences from monolingual data in Common Crawl and ParaCrawl were identified via Language-Agnostic Sentence Representation (LASER) encoders.", "### Source Data", "#### Initial Data Collection and Normalization\n\n\nMonolingual data was obtained from Common Crawl and ParaCrawl.", "#### Who are the source language producers?\n\n\nContributors to web text in Common Crawl and ParaCrawl.", "### Annotations", "#### Annotation process\n\n\nThe data was not human annotated. The metadata used to create the dataset can be found here: URL", "#### Who are the annotators?\n\n\nThe data was not human annotated. Parallel text from Common Crawl and Para Crawl monolingual data were identified automatically via LASER encoders.", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset\n\n\nThis dataset provides data for training machine learning systems for many languages that have low resources available for NLP.", "### Discussion of Biases\n\n\nBiases in the data have not been studied.", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nThe dataset is released under the terms of ODC-BY. By using this, you are also bound by the Internet Archive Terms of Use in respect of the content contained in the dataset.\n\n\nNLLB Team et al, No Language Left Behind: Scaling Human-Centered Machine Translation, Arxiv, 2022.", "### Contributions\n\n\nWe thank the AllenNLP team at AI2 for hosting and releasing this data, including Akshita Bhagia (for engineering efforts to create the huggingface dataset), and Jesse Dodge (for organizing the connection)." ]

[ "task_ids:closed-domain-qa", "annotations_creators:expert-generated", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:10k<n<100K", "source_datasets:original", "language:en", "license:other", "arxiv:1705.00754", "region:us" ]

[ "1705.00754" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThe ActivityNet Captions dataset connects videos to a series of temporally annotated sentence descriptions. Each sentence covers an unique segment of the video, describing multiple events that occur. These events may occur over very long or short periods of time and are not limited in any capacity, allowing them to co-occur. On average, each of the 20k videos contains 3.65 temporally localized sentences, resulting in a total of 100k sentences. We find that the number of sentences per video follows a relatively normal distribution. Furthermore, as the video duration increases, the number of sentences also increases. Each sentence has an average length of 13.48 words, which is also normally distributed. You can find more details of the dataset under the ActivityNet Captions Dataset section, and under supplementary materials in the paper.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_id' : 'str' unique identifier for the video\n* 'video\\_path': 'str' Path to the video file\n-'duration': 'float32' Duration of the video\n* 'captions\\_starts': 'List\\_float32' List of timestamps denoting the time at which each caption starts\n* 'captions\\_ends': 'List\\_float32' List of timestamps denoting the time at which each caption ends\n* 'en\\_captions': 'list\\_str' List of english captions describing parts of the video", "### Data Splits", "### Annotations\n\n\nQuoting ActivityNet Captions' paper: \n\n\"Each annotation task was divided into two steps: (1)\nWriting a paragraph describing all major events happening\nin the videos in a paragraph, with each sentence of the paragraph describing one event, and (2) Labeling the\nstart and end time in the video in which each sentence in the\nparagraph event occurred.\"", "### Who annotated the dataset?\n\n\nAmazon Mechnical Turk annotators", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "TAGS\n#task_ids-closed-domain-qa #annotations_creators-expert-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-10k<n<100K #source_datasets-original #language-English #license-other #arxiv-1705.00754 #region-us \n", "### Dataset Summary\n\n\nThe ActivityNet Captions dataset connects videos to a series of temporally annotated sentence descriptions. Each sentence covers an unique segment of the video, describing multiple events that occur. These events may occur over very long or short periods of time and are not limited in any capacity, allowing them to co-occur. On average, each of the 20k videos contains 3.65 temporally localized sentences, resulting in a total of 100k sentences. We find that the number of sentences per video follows a relatively normal distribution. Furthermore, as the video duration increases, the number of sentences also increases. Each sentence has an average length of 13.48 words, which is also normally distributed. You can find more details of the dataset under the ActivityNet Captions Dataset section, and under supplementary materials in the paper.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_id' : 'str' unique identifier for the video\n* 'video\\_path': 'str' Path to the video file\n-'duration': 'float32' Duration of the video\n* 'captions\\_starts': 'List\\_float32' List of timestamps denoting the time at which each caption starts\n* 'captions\\_ends': 'List\\_float32' List of timestamps denoting the time at which each caption ends\n* 'en\\_captions': 'list\\_str' List of english captions describing parts of the video", "### Data Splits", "### Annotations\n\n\nQuoting ActivityNet Captions' paper: \n\n\"Each annotation task was divided into two steps: (1)\nWriting a paragraph describing all major events happening\nin the videos in a paragraph, with each sentence of the paragraph describing one event, and (2) Labeling the\nstart and end time in the video in which each sentence in the\nparagraph event occurred.\"", "### Who annotated the dataset?\n\n\nAmazon Mechnical Turk annotators", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "task_categories:question-answering", "task_categories:visual-question-answering", "task_ids:closed-domain-qa", "annotations_creators:expert-generated", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:other", "arxiv:1604.02748", "region:us" ]

[ "1604.02748" ]

[ "en" ]

[ "### Dataset Summary\n\n\nThe Tumblr GIF (TGIF) dataset contains 100K animated GIFs and 120K sentences describing visual content of the animated GIFs. The animated GIFs have been collected from Tumblr, from randomly selected posts published between May and June of 2015. We provide the URLs of animated GIFs in this release. The sentences are collected via crowdsourcing, with a carefully designed annotation interface that ensures high quality dataset. We provide one sentence per animated GIF for the training and validation splits, and three sentences per GIF for the test split. The dataset shall be used to evaluate animated GIF/video description techniques.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_path': 'str' \"URL\n-'video\\_bytes': 'large\\_bytes' video file in bytes format\n* 'en\\_global\\_captions': 'list\\_str' List of english captions describing the entire video", "### Data Splits", "### Annotations\n\n\nQuoting TGIF paper: \n\n\"We annotated animated GIFs with natural language descriptions using the crowdsourcing service CrowdFlower.\nWe carefully designed our annotation task with various\nquality control mechanisms to ensure the sentences are both\nsyntactically and semantically of high quality.\nA total of 931 workers participated in our annotation\ntask. We allowed workers only from Australia, Canada, New Zealand, UK and USA in an effort to collect fluent descriptions from native English speakers. Figure 2 shows the\ninstructions given to the workers. Each task showed 5 animated GIFs and asked the worker to describe each with one\nsentence. To promote language style diversity, each worker\ncould rate no more than 800 images (0.7% of our corpus).\nWe paid 0.02 USD per sentence; the entire crowdsourcing\ncost less than 4K USD. We provide details of our annotation\ntask in the supplementary material.\"", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThis dataset is provided to be used for approved non-commercial research purposes. No personally identifying information is available in this dataset.", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "TAGS\n#task_categories-question-answering #task_categories-visual-question-answering #task_ids-closed-domain-qa #annotations_creators-expert-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-other #arxiv-1604.02748 #region-us \n", "### Dataset Summary\n\n\nThe Tumblr GIF (TGIF) dataset contains 100K animated GIFs and 120K sentences describing visual content of the animated GIFs. The animated GIFs have been collected from Tumblr, from randomly selected posts published between May and June of 2015. We provide the URLs of animated GIFs in this release. The sentences are collected via crowdsourcing, with a carefully designed annotation interface that ensures high quality dataset. We provide one sentence per animated GIF for the training and validation splits, and three sentences per GIF for the test split. The dataset shall be used to evaluate animated GIF/video description techniques.", "### Languages\n\n\nThe captions in the dataset are in English.\n\n\nDataset Structure\n-----------------", "### Data Fields\n\n\n* 'video\\_path': 'str' \"URL\n-'video\\_bytes': 'large\\_bytes' video file in bytes format\n* 'en\\_global\\_captions': 'list\\_str' List of english captions describing the entire video", "### Data Splits", "### Annotations\n\n\nQuoting TGIF paper: \n\n\"We annotated animated GIFs with natural language descriptions using the crowdsourcing service CrowdFlower.\nWe carefully designed our annotation task with various\nquality control mechanisms to ensure the sentences are both\nsyntactically and semantically of high quality.\nA total of 931 workers participated in our annotation\ntask. We allowed workers only from Australia, Canada, New Zealand, UK and USA in an effort to collect fluent descriptions from native English speakers. Figure 2 shows the\ninstructions given to the workers. Each task showed 5 animated GIFs and asked the worker to describe each with one\nsentence. To promote language style diversity, each worker\ncould rate no more than 800 images (0.7% of our corpus).\nWe paid 0.02 USD per sentence; the entire crowdsourcing\ncost less than 4K USD. We provide details of our annotation\ntask in the supplementary material.\"", "### Personal and Sensitive Information\n\n\nNothing specifically mentioned in the paper.\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThis dataset is provided to be used for approved non-commercial research purposes. No personally identifying information is available in this dataset.", "### Contributions\n\n\nThanks to @leot13 for adding this dataset." ]

[ "task_categories:token-classification", "task_categories:text-classification", "task_ids:multi-class-classification", "task_ids:named-entity-recognition", "task_ids:parsing", "annotations_creators:expert-generated", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:10K<n<100K", "language:en", "license:cc-by-4.0", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for sd-nlp", "## Table of Contents\n- [Dataset Card for [EMBO/sd-nlp-non-tokenized]](#dataset-card-for-dataset-name)\n - Table of Contents\n - Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n - Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n - Dataset Creation\n - Curation Rationale\n - Source Data\n - Initial Data Collection and Normalization\n - Who are the source language producers?\n - Annotations\n - Annotation process\n - Who are the annotators?\n - Personal and Sensitive Information\n - Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n - Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper:\n- Leaderboard:\n- Point of Contact: thomas.lemberger@URL, URL@URL", "### Dataset Summary\nThis dataset is based on the content of the SourceData (URL) database, which contains manually annotated figure legends written in English and extracted from scientific papers in the domain of cell and molecular biology (Liechti et al, Nature Methods, 2017, URL \nUnlike the dataset 'sd-nlp', pre-tokenized with the 'roberta-base' tokenizer, this dataset is not previously tokenized, but just splitted into words. Users can therefore use it to fine-tune other models. \nAdditional details at URL", "### Supported Tasks and Leaderboards\nTags are provided as IOB2-style tags).\n'PANELIZATION': figure captions (or figure legends) are usually composed of segments that each refer to one of several 'panels' of the full figure. Panels tend to represent results obtained with a coherent method and depicts data points that can be meaningfully compared to each other. 'PANELIZATION' provide the start (B-PANEL_START) of these segments and allow to train for recogntion of the boundary between consecutive panel lengends.\n'NER': biological and chemical entities are labeled. Specifically the following entities are tagged:\n- 'SMALL_MOLECULE': small molecules\n- 'GENEPROD': gene products (genes and proteins)\n- 'SUBCELLULAR': subcellular components\n- 'CELL': cell types and cell lines.\n- 'TISSUE': tissues and organs\n- 'ORGANISM': species\n- 'DISEASE': diseases (see limitations)\n- 'EXP_ASSAY': experimental assays\n'ROLES': the role of entities with regard to the causal hypotheses tested in the reported results. The tags are:\n- 'CONTROLLED_VAR': entities that are associated with experimental variables and that subjected to controlled and targeted perturbations.\n- 'MEASURED_VAR': entities that are associated with the variables measured and the object of the measurements.\n'BORING': entities are marked with the tag 'BORING' when they are more of descriptive value and not directly associated with causal hypotheses ('boring' is not an ideal choice of word, but it is short...). Typically, these entities are so-called 'reporter' geneproducts, entities used as common baseline across samples, or specify the context of the experiment (cellular system, species, etc...).", "### Languages\nThe text in the dataset is English.", "## Dataset Structure", "### Data Instances", "### Data Fields\n\n- 'words': 'list' of 'strings' text tokenized into words.\n- 'panel_id': ID of the panel to which the example belongs to in the SourceData database.\n- 'label_ids':\n - 'entity_types': 'list' of 'strings' for the IOB2 tags for entity type; possible value in '[\"O\", \"I-SMALL_MOLECULE\", \"B-SMALL_MOLECULE\", \"I-GENEPROD\", \"B-GENEPROD\", \"I-SUBCELLULAR\", \"B-SUBCELLULAR\", \"I-CELL\", \"B-CELL\", \"I-TISSUE\", \"B-TISSUE\", \"I-ORGANISM\", \"B-ORGANISM\", \"I-EXP_ASSAY\", \"B-EXP_ASSAY\"]'\n - 'geneprod_roles': 'list' of 'strings' for the IOB2 tags for experimental roles; values in '[\"O\", \"I-CONTROLLED_VAR\", \"B-CONTROLLED_VAR\", \"I-MEASURED_VAR\", \"B-MEASURED_VAR\"]'\n - 'boring': 'list' of 'strings' for IOB2 tags for entities unrelated to causal design; values in '[\"O\", \"I-BORING\", \"B-BORING\"]'\n - 'panel_start': 'list' of 'strings' for IOB2 tags '[\"O\", \"B-PANEL_START\"]' \n - 'small_mol_roles': 'list' of 'strings' for IOB2 tags showing whether the entity is the variable being measured or the control variable '[\"O\", \"B-CONTROLLED_VAR\", \"I-CONTROLLED_VAR\", \"B-MEASURED_VAR\", \"I-MEASURED_VAR\",]'", "### Data Splits\n\n- train:\n - features: ['words', 'labels', 'tag_mask', 'panel_id'],\n - num_rows: 50_198\n- validation:\n - features: ['words', 'labels', 'tag_mask', 'panel_id'],\n - num_rows: 5_946\n- test:\n - features: ['words', 'labels', 'tag_mask', 'panel_id'],\n - num_rows: 6_222", "## Dataset Creation", "### Curation Rationale\n\nThe dataset was built to train models for the automatic extraction of a knowledge graph based from the scientific literature. The dataset can be used to train models for text segmentation, named entity recognition and semantic role labeling.", "### Source Data", "#### Initial Data Collection and Normalization\n\nFigure legends were annotated according to the SourceData framework described in Liechti et al 2017 (Nature Methods, 2017, URL The curation tool at URL was used to segment figure legends into panel legends, tag enities, assign experiemental roles and normalize with standard identifiers (not available in this dataset). The source data was downloaded from the SourceData API (URL) on 21 Jan 2021.", "#### Who are the source language producers?\n\nThe examples are extracted from the figure legends from scientific papers in cell and molecular biology.", "### Annotations", "#### Annotation process\n\nThe annotations were produced manually with expert curators from the SourceData project (URL)", "#### Who are the annotators?\n\nCurators of the SourceData project.", "### Personal and Sensitive Information\n\nNone known.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nNot applicable.", "### Discussion of Biases\n\nThe examples are heavily biased towards cell and molecular biology and are enriched in examples from papers published in EMBO Press journals (URL)\n\nThe annotation of diseases has been added recently to the dataset. Although they appear, the number is very low and they are not consistently tagged through the entire dataset. \nWe recommend to use the diseases by filtering the examples that contain them.", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThomas Lemberger, EMBO.\nJorge Abreu Vicente, EMBO", "### Licensing Information\n\nCC BY 4.0\n\n\n\nWe are currently working on a paper to present the dataset. It is expected to be ready by 2023 spring. In the meantime, the following paper should be cited.", "### Contributions\n\nThanks to @tlemberger and @drAbreu for adding this dataset." ]

[ "TAGS\n#task_categories-token-classification #task_categories-text-classification #task_ids-multi-class-classification #task_ids-named-entity-recognition #task_ids-parsing #annotations_creators-expert-generated #language_creators-expert-generated #multilinguality-monolingual #size_categories-10K<n<100K #language-English #license-cc-by-4.0 #region-us \n", "# Dataset Card for sd-nlp", "## Table of Contents\n- [Dataset Card for [EMBO/sd-nlp-non-tokenized]](#dataset-card-for-dataset-name)\n - Table of Contents\n - Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n - Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n - Dataset Creation\n - Curation Rationale\n - Source Data\n - Initial Data Collection and Normalization\n - Who are the source language producers?\n - Annotations\n - Annotation process\n - Who are the annotators?\n - Personal and Sensitive Information\n - Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n - Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper:\n- Leaderboard:\n- Point of Contact: thomas.lemberger@URL, URL@URL", "### Dataset Summary\nThis dataset is based on the content of the SourceData (URL) database, which contains manually annotated figure legends written in English and extracted from scientific papers in the domain of cell and molecular biology (Liechti et al, Nature Methods, 2017, URL \nUnlike the dataset 'sd-nlp', pre-tokenized with the 'roberta-base' tokenizer, this dataset is not previously tokenized, but just splitted into words. Users can therefore use it to fine-tune other models. \nAdditional details at URL", "### Supported Tasks and Leaderboards\nTags are provided as IOB2-style tags).\n'PANELIZATION': figure captions (or figure legends) are usually composed of segments that each refer to one of several 'panels' of the full figure. Panels tend to represent results obtained with a coherent method and depicts data points that can be meaningfully compared to each other. 'PANELIZATION' provide the start (B-PANEL_START) of these segments and allow to train for recogntion of the boundary between consecutive panel lengends.\n'NER': biological and chemical entities are labeled. Specifically the following entities are tagged:\n- 'SMALL_MOLECULE': small molecules\n- 'GENEPROD': gene products (genes and proteins)\n- 'SUBCELLULAR': subcellular components\n- 'CELL': cell types and cell lines.\n- 'TISSUE': tissues and organs\n- 'ORGANISM': species\n- 'DISEASE': diseases (see limitations)\n- 'EXP_ASSAY': experimental assays\n'ROLES': the role of entities with regard to the causal hypotheses tested in the reported results. The tags are:\n- 'CONTROLLED_VAR': entities that are associated with experimental variables and that subjected to controlled and targeted perturbations.\n- 'MEASURED_VAR': entities that are associated with the variables measured and the object of the measurements.\n'BORING': entities are marked with the tag 'BORING' when they are more of descriptive value and not directly associated with causal hypotheses ('boring' is not an ideal choice of word, but it is short...). Typically, these entities are so-called 'reporter' geneproducts, entities used as common baseline across samples, or specify the context of the experiment (cellular system, species, etc...).", "### Languages\nThe text in the dataset is English.", "## Dataset Structure", "### Data Instances", "### Data Fields\n\n- 'words': 'list' of 'strings' text tokenized into words.\n- 'panel_id': ID of the panel to which the example belongs to in the SourceData database.\n- 'label_ids':\n - 'entity_types': 'list' of 'strings' for the IOB2 tags for entity type; possible value in '[\"O\", \"I-SMALL_MOLECULE\", \"B-SMALL_MOLECULE\", \"I-GENEPROD\", \"B-GENEPROD\", \"I-SUBCELLULAR\", \"B-SUBCELLULAR\", \"I-CELL\", \"B-CELL\", \"I-TISSUE\", \"B-TISSUE\", \"I-ORGANISM\", \"B-ORGANISM\", \"I-EXP_ASSAY\", \"B-EXP_ASSAY\"]'\n - 'geneprod_roles': 'list' of 'strings' for the IOB2 tags for experimental roles; values in '[\"O\", \"I-CONTROLLED_VAR\", \"B-CONTROLLED_VAR\", \"I-MEASURED_VAR\", \"B-MEASURED_VAR\"]'\n - 'boring': 'list' of 'strings' for IOB2 tags for entities unrelated to causal design; values in '[\"O\", \"I-BORING\", \"B-BORING\"]'\n - 'panel_start': 'list' of 'strings' for IOB2 tags '[\"O\", \"B-PANEL_START\"]' \n - 'small_mol_roles': 'list' of 'strings' for IOB2 tags showing whether the entity is the variable being measured or the control variable '[\"O\", \"B-CONTROLLED_VAR\", \"I-CONTROLLED_VAR\", \"B-MEASURED_VAR\", \"I-MEASURED_VAR\",]'", "### Data Splits\n\n- train:\n - features: ['words', 'labels', 'tag_mask', 'panel_id'],\n - num_rows: 50_198\n- validation:\n - features: ['words', 'labels', 'tag_mask', 'panel_id'],\n - num_rows: 5_946\n- test:\n - features: ['words', 'labels', 'tag_mask', 'panel_id'],\n - num_rows: 6_222", "## Dataset Creation", "### Curation Rationale\n\nThe dataset was built to train models for the automatic extraction of a knowledge graph based from the scientific literature. The dataset can be used to train models for text segmentation, named entity recognition and semantic role labeling.", "### Source Data", "#### Initial Data Collection and Normalization\n\nFigure legends were annotated according to the SourceData framework described in Liechti et al 2017 (Nature Methods, 2017, URL The curation tool at URL was used to segment figure legends into panel legends, tag enities, assign experiemental roles and normalize with standard identifiers (not available in this dataset). The source data was downloaded from the SourceData API (URL) on 21 Jan 2021.", "#### Who are the source language producers?\n\nThe examples are extracted from the figure legends from scientific papers in cell and molecular biology.", "### Annotations", "#### Annotation process\n\nThe annotations were produced manually with expert curators from the SourceData project (URL)", "#### Who are the annotators?\n\nCurators of the SourceData project.", "### Personal and Sensitive Information\n\nNone known.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nNot applicable.", "### Discussion of Biases\n\nThe examples are heavily biased towards cell and molecular biology and are enriched in examples from papers published in EMBO Press journals (URL)\n\nThe annotation of diseases has been added recently to the dataset. Although they appear, the number is very low and they are not consistently tagged through the entire dataset. \nWe recommend to use the diseases by filtering the examples that contain them.", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThomas Lemberger, EMBO.\nJorge Abreu Vicente, EMBO", "### Licensing Information\n\nCC BY 4.0\n\n\n\nWe are currently working on a paper to present the dataset. It is expected to be ready by 2023 spring. In the meantime, the following paper should be cited.", "### Contributions\n\nThanks to @tlemberger and @drAbreu for adding this dataset." ]

[ "task_categories:text-classification", "task_ids:sentiment-analysis", "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:1K<n<10K", "source_datasets:original", "language:yo", "license:unknown", "movie reviews", "nollywood", "arxiv:2204.09711", "region:us" ]

[ "2204.09711" ]

[ "yo" ]

[ "### Dataset Summary\n\n\nYOSM is the first Yorùbá sentiment corpus for Nollywood movie reviews. The reviews were collected from movie reviews websites - IMDB, Rotten Tomatoes, LetterboxD, Cinemapointer, and Nollyrated.", "### Languages\n\n\nYorùbá (ISO 639-1: yo) - the third most spoken indigenous African language with over 50 million speakers.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nAn instance consists of a movie review and the corresponding class label.", "### Data Fields\n\n\n* 'yo\\_review': A movie review in Yorùbá\n* 'sentiment': The label describing the sentiment of the movie review.", "### Data Splits\n\n\nThe YOSM dataset has 3 splits: *train*, *dev*, and *test*. Below are the statistics for Version 3.0.0 of the dataset.", "### Data Splits\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "TAGS\n#task_categories-text-classification #task_ids-sentiment-analysis #annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-Yoruba #license-unknown #movie reviews #nollywood #arxiv-2204.09711 #region-us \n", "### Dataset Summary\n\n\nYOSM is the first Yorùbá sentiment corpus for Nollywood movie reviews. The reviews were collected from movie reviews websites - IMDB, Rotten Tomatoes, LetterboxD, Cinemapointer, and Nollyrated.", "### Languages\n\n\nYorùbá (ISO 639-1: yo) - the third most spoken indigenous African language with over 50 million speakers.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nAn instance consists of a movie review and the corresponding class label.", "### Data Fields\n\n\n* 'yo\\_review': A movie review in Yorùbá\n* 'sentiment': The label describing the sentiment of the movie review.", "### Data Splits\n\n\nThe YOSM dataset has 3 splits: *train*, *dev*, and *test*. Below are the statistics for Version 3.0.0 of the dataset.", "### Data Splits\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "license:apache-2.0", "region:us" ]

[]

[]

[]

[ "TAGS\n#license-apache-2.0 #region-us \n" ]

[ "language:ar", "license:apache-2.0", "region:us" ]

[]

[ "ar" ]

[ "# arabench\n\n- Dataset uid: 'arabench'", "### Description\n\nAraBench is an evaluation suite for dialectal Arabic to English machine translation. AraBench offers 4 coarse, 15 fine-grained and 25 city-level dialect categories, belonging to diverse genres, such as media, chat, religion and travel with varying level of dialectness.", "### Homepage\n\nURL", "### Licensing\n\n- open license\n- cc-by-4.0: Creative Commons Attribution 4.0 International", "### Speaker Locations\n\n- Northern Africa\n- Western Asia\n- Algeria\n- Egypt\n- Morocco\n- Jordan\n- Sudan\n- Tunisia\n- Lebanon\n- Libya\n- Iraq\n- Qatar\n- Yemen\n- Oman\n- Saudi Arabia\n- Syria\n- Palestine", "### Sizes\n\n- 0.0018 % of total\n- 0.0165 % of ar", "### BigScience processing steps", "#### Filters applied to: ar\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300" ]

[ "TAGS\n#language-Arabic #license-apache-2.0 #region-us \n", "# arabench\n\n- Dataset uid: 'arabench'", "### Description\n\nAraBench is an evaluation suite for dialectal Arabic to English machine translation. AraBench offers 4 coarse, 15 fine-grained and 25 city-level dialect categories, belonging to diverse genres, such as media, chat, religion and travel with varying level of dialectness.", "### Homepage\n\nURL", "### Licensing\n\n- open license\n- cc-by-4.0: Creative Commons Attribution 4.0 International", "### Speaker Locations\n\n- Northern Africa\n- Western Asia\n- Algeria\n- Egypt\n- Morocco\n- Jordan\n- Sudan\n- Tunisia\n- Lebanon\n- Libya\n- Iraq\n- Qatar\n- Yemen\n- Oman\n- Saudi Arabia\n- Syria\n- Palestine", "### Sizes\n\n- 0.0018 % of total\n- 0.0165 % of ar", "### BigScience processing steps", "#### Filters applied to: ar\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300" ]

[ "language:ar", "license:gpl-2.0", "region:us" ]

[]

[ "ar" ]

[ "# labr\n\n- Dataset uid: 'labr'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0076 % of total\n- 0.0701 % of ar", "### BigScience processing steps", "#### Filters applied to: ar\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300" ]

[ "TAGS\n#language-Arabic #license-gpl-2.0 #region-us \n", "# labr\n\n- Dataset uid: 'labr'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0076 % of total\n- 0.0701 % of ar", "### BigScience processing steps", "#### Filters applied to: ar\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300" ]

[ "language:ar", "license:cc-by-sa-3.0", "region:us" ]

[]

[ "ar" ]

[ "# wikinews_filtered\n\n- Dataset uid: 'wikinews_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0307 % of total\n- 0.0701 % of ar\n- 0.3036 % of pt\n- 0.0271 % of en\n- 0.0405 % of fr\n- 0.2119 % of indic-ta\n- 0.0081 % of zh\n- 0.0510 % of es\n- 0.0725 % of ca", "### BigScience processing steps", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ar\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_pt\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: en\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_en\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-ta\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: zh\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_zhs\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_es\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ca\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024" ]

[ "TAGS\n#language-Arabic #license-cc-by-sa-3.0 #region-us \n", "# wikinews_filtered\n\n- Dataset uid: 'wikinews_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0307 % of total\n- 0.0701 % of ar\n- 0.3036 % of pt\n- 0.0271 % of en\n- 0.0405 % of fr\n- 0.2119 % of indic-ta\n- 0.0081 % of zh\n- 0.0510 % of es\n- 0.0725 % of ca", "### BigScience processing steps", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ar\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_pt\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: en\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_en\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-ta\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: zh\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_zhs\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_es\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ca\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024" ]

[ "language:ar", "license:cc-by-sa-3.0", "region:us" ]

[]

[ "ar" ]

[ "# wikiquote_filtered\n\n- Dataset uid: 'wikiquote_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0462 % of total\n- 0.1697 % of en\n- 0.0326 % of fr\n- 0.0216 % of ar\n- 0.0066 % of zh\n- 0.0833 % of pt\n- 0.0357 % of es\n- 0.0783 % of indic-ta\n- 0.0361 % of indic-hi\n- 0.0518 % of ca\n- 0.0405 % of vi\n- 0.0834 % of indic-ml\n- 0.0542 % of indic-te\n- 0.1172 % of indic-gu\n- 0.0634 % of indic-kn\n- 0.0539 % of id\n- 0.0454 % of indic-ur\n- 0.0337 % of indic-mr\n- 0.0347 % of eu", "### BigScience processing steps", "#### Filters applied to: en\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_en\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_fr\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ar\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: zh\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_zhs\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_pt\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_es\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-ta\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-hi\n- dedup_template_soft\n- filter_small_docs_bytes_300", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ca\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: vi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_vi\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ml\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-ml\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-te\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-te\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-gu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-gu\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-kn\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-kn\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: id\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_id\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ur\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-mr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-mr\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: eu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_eu\n- dedup_template_soft\n- replace_newline_with_space" ]

[ "TAGS\n#language-Arabic #license-cc-by-sa-3.0 #region-us \n", "# wikiquote_filtered\n\n- Dataset uid: 'wikiquote_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0462 % of total\n- 0.1697 % of en\n- 0.0326 % of fr\n- 0.0216 % of ar\n- 0.0066 % of zh\n- 0.0833 % of pt\n- 0.0357 % of es\n- 0.0783 % of indic-ta\n- 0.0361 % of indic-hi\n- 0.0518 % of ca\n- 0.0405 % of vi\n- 0.0834 % of indic-ml\n- 0.0542 % of indic-te\n- 0.1172 % of indic-gu\n- 0.0634 % of indic-kn\n- 0.0539 % of id\n- 0.0454 % of indic-ur\n- 0.0337 % of indic-mr\n- 0.0347 % of eu", "### BigScience processing steps", "#### Filters applied to: en\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_en\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_fr\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ar\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: zh\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_zhs\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_pt\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_es\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-ta\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-hi\n- dedup_template_soft\n- filter_small_docs_bytes_300", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ca\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: vi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_vi\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ml\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-ml\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-te\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-te\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-gu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-gu\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-kn\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-kn\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: id\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_id\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ur\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-mr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-mr\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: eu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_eu\n- dedup_template_soft\n- replace_newline_with_space" ]

[ "language:ar", "license:cc-by-sa-3.0", "region:us" ]

[]

[ "ar" ]

[ "# wikiversity_filtered\n\n- Dataset uid: 'wikiversity_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0367 % of total\n- 0.1050 % of en\n- 0.1178 % of fr\n- 0.1231 % of pt\n- 0.0072 % of zh\n- 0.0393 % of es\n- 0.0076 % of ar\n- 0.0069 % of indic-hi", "### BigScience processing steps", "#### Filters applied to: en\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_en\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_fr\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_pt\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: zh\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_zhs\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_es\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ar\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-hi\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300" ]

[ "TAGS\n#language-Arabic #license-cc-by-sa-3.0 #region-us \n", "# wikiversity_filtered\n\n- Dataset uid: 'wikiversity_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 0.0367 % of total\n- 0.1050 % of en\n- 0.1178 % of fr\n- 0.1231 % of pt\n- 0.0072 % of zh\n- 0.0393 % of es\n- 0.0076 % of ar\n- 0.0069 % of indic-hi", "### BigScience processing steps", "#### Filters applied to: en\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_en\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_fr\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_pt\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: zh\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_zhs\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_es\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_1024", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_ar\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- filter_remove_empty_docs\n- split_sentences_indic-hi\n- dedup_template_soft\n- replace_newline_with_space\n- filter_small_docs_bytes_300" ]

[ "language:ca", "license:cc0-1.0", "region:us" ]

[]

[ "ca" ]

[ "# Catalan Government Crawling\n\n- Dataset uid: 'catalan_government_crawling'", "### Description\n\nThe Catalan Government Crawling Corpus is a 39-million-token web corpus of Catalan built from the web. It has been obtained by crawling the .gencat domain and subdomains, belonging to the Catalan Government during September and October 2020. It consists of 39.117.909 tokens, 1.565.433 sentences and 71.043 documents. Documents are separated by single new lines. It is a subcorpus of the Catalan Textual Corpus.", "### Homepage\n\nURL", "### Licensing\n\n- open license\n- cc0-1.0: Creative Commons Zero v1.0 Universal", "### Speaker Locations\n\n- Southern Europe\n- Spain", "### Sizes\n\n- 0.0219 % of total\n- 1.8426 % of ca", "### BigScience processing steps", "#### Filters applied to: ca\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024" ]

[ "TAGS\n#language-Catalan #license-cc0-1.0 #region-us \n", "# Catalan Government Crawling\n\n- Dataset uid: 'catalan_government_crawling'", "### Description\n\nThe Catalan Government Crawling Corpus is a 39-million-token web corpus of Catalan built from the web. It has been obtained by crawling the .gencat domain and subdomains, belonging to the Catalan Government during September and October 2020. It consists of 39.117.909 tokens, 1.565.433 sentences and 71.043 documents. Documents are separated by single new lines. It is a subcorpus of the Catalan Textual Corpus.", "### Homepage\n\nURL", "### Licensing\n\n- open license\n- cc0-1.0: Creative Commons Zero v1.0 Universal", "### Speaker Locations\n\n- Southern Europe\n- Spain", "### Sizes\n\n- 0.0219 % of total\n- 1.8426 % of ca", "### BigScience processing steps", "#### Filters applied to: ca\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024" ]

[ "language:ar", "license:cc-by-sa-3.0", "region:us" ]

[]

[ "ar" ]

[ "# wikisource_filtered\n\n- Dataset uid: 'wikisource_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 2.6306 % of total\n- 12.7884 % of fr\n- 19.8886 % of indic-bn\n- 20.9966 % of indic-ta\n- 2.3478 % of ar\n- 4.7068 % of indic-hi\n- 18.0998 % of indic-te\n- 1.7155 % of es\n- 19.4800 % of indic-kn\n- 9.1737 % of indic-ml\n- 17.1771 % of indic-mr\n- 17.1870 % of indic-gu\n- 70.3687 % of indic-as\n- 1.0165 % of pt\n- 7.8642 % of indic-pa\n- 1.3501 % of vi\n- 4.9411 % of indic-or\n- 0.5307 % of ca\n- 2.3593 % of id\n- 1.5928 % of eu", "### BigScience processing steps", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-bn\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-te\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-kn\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- remove_wiki_mojibake\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ml\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-mr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-gu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-as\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-pa\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: vi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-or\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: id\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: eu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs" ]

[ "TAGS\n#language-Arabic #license-cc-by-sa-3.0 #region-us \n", "# wikisource_filtered\n\n- Dataset uid: 'wikisource_filtered'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 2.6306 % of total\n- 12.7884 % of fr\n- 19.8886 % of indic-bn\n- 20.9966 % of indic-ta\n- 2.3478 % of ar\n- 4.7068 % of indic-hi\n- 18.0998 % of indic-te\n- 1.7155 % of es\n- 19.4800 % of indic-kn\n- 9.1737 % of indic-ml\n- 17.1771 % of indic-mr\n- 17.1870 % of indic-gu\n- 70.3687 % of indic-as\n- 1.0165 % of pt\n- 7.8642 % of indic-pa\n- 1.3501 % of vi\n- 4.9411 % of indic-or\n- 0.5307 % of ca\n- 2.3593 % of id\n- 1.5928 % of eu", "### BigScience processing steps", "#### Filters applied to: fr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-bn\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-te\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: es\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: indic-kn\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- remove_wiki_mojibake\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ml\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-mr\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-gu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-as\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs", "#### Filters applied to: pt\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-pa\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: vi\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-or\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: id\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: eu\n\n- filter_wiki_user_titles\n- filter_wiki_non_text_type\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs" ]

[ "language:ca", "license:cc-by-4.0", "region:us" ]

[]

[ "ca" ]

[ "# Enriched CONLLU Ancora for ML training\n\n- Dataset uid: 'enriched_conllu_ancora_for_ml_training'", "### Description\n\nThis is an enriched version for Machine Learning purposes of the CONLLU adaptation of AnCora corpus .\n\nThis version of the corpus was developed by BSC TeMU as part of the AINA project, and has been used to do multi-task learning for the Catalan language Spacy 3.0 models.", "### Homepage\n\nURL", "### Licensing\n\n- cc-by-4.0: Creative Commons Attribution 4.0 International", "### Speaker Locations\n\n- Spain", "### Sizes\n\n- 0.0000 % of total\n- 0.0000 % of ca", "### BigScience processing steps", "#### Filters applied to: ca\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024" ]

[ "TAGS\n#language-Catalan #license-cc-by-4.0 #region-us \n", "# Enriched CONLLU Ancora for ML training\n\n- Dataset uid: 'enriched_conllu_ancora_for_ml_training'", "### Description\n\nThis is an enriched version for Machine Learning purposes of the CONLLU adaptation of AnCora corpus .\n\nThis version of the corpus was developed by BSC TeMU as part of the AINA project, and has been used to do multi-task learning for the Catalan language Spacy 3.0 models.", "### Homepage\n\nURL", "### Licensing\n\n- cc-by-4.0: Creative Commons Attribution 4.0 International", "### Speaker Locations\n\n- Spain", "### Sizes\n\n- 0.0000 % of total\n- 0.0000 % of ca", "### BigScience processing steps", "#### Filters applied to: ca\n\n- dedup_document\n- dedup_template_soft\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024" ]

[ "language:ar", "license:cc-by-sa-3.0", "region:us" ]

[]

[ "ar" ]

[ "# wikipedia\n\n- Dataset uid: 'wikipedia'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 3.2299 % of total\n- 4.2071 % of en\n- 5.6773 % of ar\n- 3.3416 % of fr\n- 5.2815 % of es\n- 12.4852 % of ca\n- 0.4288 % of zh\n- 0.4286 % of zh\n- 5.4743 % of indic-bn\n- 8.9062 % of indic-ta\n- 21.3313 % of indic-te\n- 4.4845 % of pt\n- 4.0493 % of indic-hi\n- 11.3163 % of indic-ml\n- 22.5300 % of indic-ur\n- 4.4902 % of vi\n- 16.9916 % of indic-kn\n- 24.7820 % of eu\n- 11.6241 % of indic-mr\n- 9.8749 % of id\n- 9.3489 % of indic-pa\n- 9.4767 % of indic-gu\n- 24.1132 % of indic-as\n- 5.3309 % of indic-or", "### BigScience processing steps", "#### Filters applied to: en\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: fr\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: es\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: zh", "#### Filters applied to: zh", "#### Filters applied to: indic-bn\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-te\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: pt\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ml\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ur\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: vi\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-kn\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: eu\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs", "#### Filters applied to: indic-mr\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: id\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-pa\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-gu\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-as\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs", "#### Filters applied to: indic-or\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs" ]

[ "TAGS\n#language-Arabic #license-cc-by-sa-3.0 #region-us \n", "# wikipedia\n\n- Dataset uid: 'wikipedia'", "### Description", "### Homepage", "### Licensing", "### Speaker Locations", "### Sizes\n\n- 3.2299 % of total\n- 4.2071 % of en\n- 5.6773 % of ar\n- 3.3416 % of fr\n- 5.2815 % of es\n- 12.4852 % of ca\n- 0.4288 % of zh\n- 0.4286 % of zh\n- 5.4743 % of indic-bn\n- 8.9062 % of indic-ta\n- 21.3313 % of indic-te\n- 4.4845 % of pt\n- 4.0493 % of indic-hi\n- 11.3163 % of indic-ml\n- 22.5300 % of indic-ur\n- 4.4902 % of vi\n- 16.9916 % of indic-kn\n- 24.7820 % of eu\n- 11.6241 % of indic-mr\n- 9.8749 % of id\n- 9.3489 % of indic-pa\n- 9.4767 % of indic-gu\n- 24.1132 % of indic-as\n- 5.3309 % of indic-or", "### BigScience processing steps", "#### Filters applied to: en\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: ar\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: fr\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: es\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: ca\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_1024", "#### Filters applied to: zh", "#### Filters applied to: zh", "#### Filters applied to: indic-bn\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ta\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-te\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: pt\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-hi\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ml\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-ur\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: vi\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-kn\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: eu\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs", "#### Filters applied to: indic-mr\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: id\n\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-pa\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-gu\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs\n- filter_small_docs_bytes_300", "#### Filters applied to: indic-as\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs", "#### Filters applied to: indic-or\n\n- filter_wiki_user_titles\n- dedup_document\n- filter_remove_empty_docs" ]