gilkeyio commited on
Commit
757c9f7
1 Parent(s): 4abc9d0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +39 -86
README.md CHANGED
@@ -78,7 +78,9 @@ size_categories:
78
  - 100K<n<1M
79
  ---
80
 
81
- # Dataset Card for Dataset Name
 
 
82
 
83
  Librispeech with alignments generated by the [Montreal Forced Aligner](https://montreal-forced-aligner.readthedocs.io/en/latest/). The original alignments in TextGrid format can be found [here](https://zenodo.org/records/2619474)
84
 
@@ -87,140 +89,91 @@ Librispeech with alignments generated by the [Montreal Forced Aligner](https://m
87
 
88
  ### Dataset Description
89
 
90
- <!-- Provide a longer summary of what this dataset is. -->
91
 
 
92
 
93
- - **Curated by:** [More Information Needed]
94
- - **Funded by [optional]:** [More Information Needed]
95
- - **Shared by [optional]:** [More Information Needed]
96
- - **Language(s) (NLP):** [More Information Needed]
97
- - **License:** [More Information Needed]
98
 
99
- ### Dataset Sources [optional]
 
 
 
 
100
 
101
- <!-- Provide the basic links for the dataset. -->
102
 
103
- - **Repository:** [More Information Needed]
104
- - **Paper [optional]:** [More Information Needed]
105
- - **Demo [optional]:** [More Information Needed]
106
 
107
  ## Uses
108
 
109
- <!-- Address questions around how the dataset is intended to be used. -->
110
-
111
  ### Direct Use
112
 
113
- <!-- This section describes suitable use cases for the dataset. -->
114
-
115
- [More Information Needed]
116
 
117
  ### Out-of-Scope Use
118
 
119
- <!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
120
-
121
- [More Information Needed]
122
 
123
  ## Dataset Structure
124
 
125
- <!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
126
 
127
- [More Information Needed]
128
 
129
  ## Dataset Creation
130
 
131
  ### Curation Rationale
132
 
133
- <!-- Motivation for the creation of this dataset. -->
134
-
135
- [More Information Needed]
136
-
137
- ### Source Data
138
 
139
- <!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
140
 
141
  #### Data Collection and Processing
142
 
143
- <!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
144
 
145
- [More Information Needed]
146
 
147
- #### Who are the source data producers?
148
 
149
- <!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
150
 
151
- [More Information Needed]
152
 
153
- ### Annotations [optional]
154
-
155
- <!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
156
-
157
- #### Annotation process
158
-
159
- <!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
160
-
161
- [More Information Needed]
162
 
163
  #### Who are the annotators?
164
 
165
- <!-- This section describes the people or systems who created the annotations. -->
166
-
167
- [More Information Needed]
168
 
169
  #### Personal and Sensitive Information
170
 
171
- <!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
172
-
173
- [More Information Needed]
174
 
175
  ## Bias, Risks, and Limitations
176
 
177
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
178
-
179
- [More Information Needed]
180
 
181
  ### Recommendations
182
 
183
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
184
 
185
- Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
186
 
187
- ## Citation [optional]
188
-
189
- <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
190
-
191
- **BibTeX for the original Librispeech:**
192
  ```
193
- @inproceedings{panayotov2015librispeech,
194
  title={Librispeech: an ASR corpus based on public domain audio books},
195
- author={Panayotov, Vassil and Chen, Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
196
- booktitle={Acoustics, Speech and Signal Processing (ICASSP), 2015 IEEE International Conference on},
197
- pages={5206--5210},
198
- year={2015},
199
- organization={IEEE}
200
  }
201
  ```
202
 
203
- [More Information Needed]
204
-
205
  **APA for Alignments:**
206
  ```
207
- Loren Lugosch. (2019). LibriSpeech Alignments (1.0). Zenodo. https://doi.org/10.5281/zenodo.2619474
208
- ```
209
-
210
- ## Glossary [optional]
211
-
212
- <!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
213
-
214
- [More Information Needed]
215
-
216
- ## More Information [optional]
217
-
218
- [More Information Needed]
219
-
220
- ## Dataset Card Authors [optional]
221
-
222
- [More Information Needed]
223
-
224
- ## Dataset Card Contact
225
-
226
- [More Information Needed]
 
78
  - 100K<n<1M
79
  ---
80
 
81
+
82
+
83
+ # Dataset Card for Librispeech Alignments
84
 
85
  Librispeech with alignments generated by the [Montreal Forced Aligner](https://montreal-forced-aligner.readthedocs.io/en/latest/). The original alignments in TextGrid format can be found [here](https://zenodo.org/records/2619474)
86
 
 
89
 
90
  ### Dataset Description
91
 
92
+ Librispeech is a corpus of read English speech, designed for training and evaluating automatic speech recognition (ASR) systems. The dataset contains 1000 hours of 16kHz read English speech derived from audiobooks.
93
 
94
+ The Montreal Forced Aligner (MFA) was used to generate word and phoneme level alignments for the Librispeech dataset.
95
 
 
 
 
 
 
96
 
97
+ - **Curated by:** Vassil Panayotov, Guoguo Chen, Daniel Povey, Sanjeev Khudanpur
98
+ - **Funded by:** DARPA LORELEI
99
+ - **Shared by:** Loren Lugosch
100
+ - **Language(s) (NLP):** English
101
+ - **License:** Creative Commons Attribution 4.0 International License
102
 
103
+ ### Dataset Sources
104
 
105
+ - **Repository:** https://www.openslr.org/12
106
+ - **Paper:** https://arxiv.org/abs/1512.02595
107
+ - **Alignments:** https://zenodo.org/record/2619474
108
 
109
  ## Uses
110
 
 
 
111
  ### Direct Use
112
 
113
+ The Librispeech dataset can be used to train and evaluate ASR systems. The alignments allow for forced alignment techniques.
 
 
114
 
115
  ### Out-of-Scope Use
116
 
117
+ The dataset only contains read speech, so may not perform as well on spontaneous conversational speech.
 
 
118
 
119
  ## Dataset Structure
120
 
121
+ The dataset contains 1000 hours of segmented read English speech from audiobooks. There are three subsets: 100 hours (train-clean-100), 360 hours (train-clean-360) and 500 hours (train-other-500).
122
 
123
+ The alignments connect the audio to the reference text transcripts on word and phoneme level.
124
 
125
  ## Dataset Creation
126
 
127
  ### Curation Rationale
128
 
129
+ Librispeech was created to further speech recognition research and to benchmark progress in the field.
 
 
 
 
130
 
131
+ ### Source Data
132
 
133
  #### Data Collection and Processing
134
 
135
+ The audio and reference texts were sourced from read English audiobooks in the LibriVox project. The data was segmented, filtered and prepared for speech recognition.
136
 
137
+ #### Who are the source data producers?
138
 
139
+ The audiobooks are read by volunteers for the LibriVox project. Information about the readers is available in the LibriVox catalog.
140
 
141
+ ### Annotations
142
 
143
+ #### Annotation process
144
 
145
+ The Montreal Forced Aligner was used to create word and phoneme level alignments between the audio and reference texts. The aligner is based on Kaldi.
 
 
 
 
 
 
 
 
146
 
147
  #### Who are the annotators?
148
 
149
+ The alignments were generated automatically by the Montreal Forced Aligner and shared by Loren Lugosch.
 
 
150
 
151
  #### Personal and Sensitive Information
152
 
153
+ The data contains read speech and transcripts. No personal or sensitive information expected.
 
 
154
 
155
  ## Bias, Risks, and Limitations
156
 
157
+ The dataset contains only read speech from published books, not natural conversational speech. Performance on other tasks may be reduced.
 
 
158
 
159
  ### Recommendations
160
 
161
+ Users should understand that the alignments may contain errors and account for this in applications.
162
 
163
+ ## Citation
164
 
165
+ **BibTeX for Librispeech:**
 
 
 
 
166
  ```
167
+ @inproceedings{panayotov2015librispeech,
168
  title={Librispeech: an ASR corpus based on public domain audio books},
169
+ author={Panayotov, Vassil and Chen, Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
170
+ booktitle={ICASSP},
171
+ year={2015},
172
+ organization={IEEE}
 
173
  }
174
  ```
175
 
 
 
176
  **APA for Alignments:**
177
  ```
178
+ Lugosch, L. (2019). LibriSpeech Alignments (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.2619474
179
+ ```