wikihow_es / README.md
mapama247's picture
Convert dataset to Parquet
fb77dc6 verified
|
raw
history blame
6.19 kB
metadata
language: es
license: cc-by-nc-sa-3.0
multilinguality: monolingual
size_categories: 1K<n<10K
task_categories:
  - text-classification
  - question-answering
  - conversational
  - summarization
pretty_name: WikiHow-ES
tags:
  - Spanish
  - WikiHow
  - Wiki Articles
  - Tutorials
  - Step-By-Step
  - Instruction Tuning
dataset_info:
  config_name: all
  features:
    - name: category
      dtype: string
    - name: question
      dtype: string
    - name: introduction
      dtype: string
    - name: answers
      sequence: string
    - name: short_answers
      sequence: string
    - name: url
      dtype: string
    - name: num_answers
      dtype: int32
    - name: num_refs
      dtype: int32
    - name: expert_author
      dtype: bool
  splits:
    - name: train
      num_bytes: 70513673
      num_examples: 7380
  download_size: 38605450
  dataset_size: 70513673
configs:
  - config_name: all
    data_files:
      - split: train
        path: all/train-*
    default: true

Dataset Summary

Articles retrieved from the Spanish WikiHow website on September 2023.

Each article contains a tutorial about a specific topic. The format is always a "How to" question followed by a detailed step-by-step explanation. In some cases, the response includes several methods.

The main idea is to use this data for instruction tuning of Spanish LLMs, but given its nature it could also be used for other tasks such as text classification or summarization.

Languages

  • Spanish (ES)

Usage

To load the full dataset:

from datasets import load_dataset

all_articles = load_dataset("mapama247/wikihow_es", trust_remote_code=True)
print(all_articles.num_rows) # output: {'train': 7380}

To load only examples from a specific category:

from datasets import load_dataset

sports_articles = load_dataset("mapama247/wikihow_es", "deportes")
print(sports_articles.num_rows) # output: {'train': 201}

List of available categories, with the repective number of examples:

computadoras-y-electrónica       821
salud                            804
pasatiempos                      729
cuidado-y-estilo-personal        724
carreras-y-educación             564
en-la-casa-y-el-jardín           496
finanzas-y-negocios              459
comida-y-diversión               454
relaciones                       388
mascotas-y-animales              338
filosofía-y-religión             264
arte-y-entretenimiento           254
en-el-trabajo                    211
adolescentes                     201
deportes                         201
vida-familiar                    147
viajes                           139
automóviles-y-otros-vehículos    100
días-de-fiesta-y-tradiciones      86

Supported Tasks

This dataset can be used to train a model for...

  • instruction-tuning
  • text-classification
  • question-answering
  • conversational
  • summarization

Dataset Structure

Data Instances

{
    'category': str,
    'question': str,
    'introduction': str,
    'answers': List[str],
    'short_answers': List[str],
    'url': str,
    'num_answers': int,
    'num_refs': int,
    'expert_author': bool,
}

Data Fields

  • category: The category (from this list) to which the example belongs to.
  • label: Numerical representation of the category, for text classification purposes.
  • question: The article's title, which always starts with "¿Cómo ...".
  • introduction: Introductory text that precedes the step-by-step explanation.
  • answers: List of complete answers, with the full explanation of each step.
  • short_answers: List of shorter answers that only contain one-sentence steps.
  • num_answers: The number of alternative answers provided (e.g. length of answers).
  • num_ref: Number of references provided in the article.
  • expert_authors: Whether the article's author claims to be an expert on the topic or not.
  • url: The URL address of the original article.

Data Splits

There is only one split (train) that contains a total of 7,380 examples.

Dataset Creation

Curation Rationale

This dataset was created for language model alignment to end tasks and user preferences.

Source Data

How-To questions with detailed step-by-step answers, retrieved from the WikiHow website.

Data Collection and Normalization

All articles available in September 2023 were extracted, no filters applied.

Along with the article's content, some metadata was retrieved as well.

Source language producers

WikiHow users. All the content is human-generated.

Personal and Sensitive Information

The data does not include personal or sensitive information.

Considerations

Social Impact

The Spanish community can benefit from the high-quality data provided by this dataset.

Bias

No post-processing steps have been applied to mitigate potential social biases.

Additional Information

Curators

Marc Pàmes @ Barcelona Supercomputing Center.

License

This dataset is licensed under a Creative Commons CC BY-NC-SA 3.0 license.

Quote from WikiHow's Terms of Use:

All text posted by Users to the Service is sub-licensed by wikiHow to other Users under a Creative Commons license as provided herein. The Creative Commons license allows such user generated text content to be used freely for personal, non-commercial purposes, so long as it is used and attributed to the original author as specified under the terms of the license. Allowing free republication of our articles helps wikiHow achieve its mission by providing instruction on solving the problems of everyday life to more people for free. In order to support this goal, wikiHow hereby grants each User of the Service a license to all text content that Users contribute to the Service under the terms and conditions of a Creative Commons CC BY-NC-SA 3.0 License. Please be sure to read the terms of the license carefully. You continue to own all right, title, and interest in and to your User Content, and you are free to distribute it as you wish, whether for commercial or non-commercial purposes.