projecte-aina/viquiquad

ViquiQuAD: An Extractive QA Dataset for Catalan from Wikipedia

Dataset Description

• —Homepage: Zenodo
• —Paper: Are Multilingual Models the Best Choice for Moderately Under-resourced Languages? A Comprehensive Assessment for Catalan
• —Point of Contact: langtech@bsc.es

Dataset Summary

ViquiQuAD is an extractive Question Answering dataset for Catalan, built from the Catalan Wikipedia (Viquipèdia).

• —3,111 contexts extracted from 597 high-quality, original (non-translated) articles.
• —For each context, 1 to 5 questions were created with their corresponding answers.
• —Total: 15,153 question–answer pairs.

This dataset can be used to fine-tune and evaluate extractive QA models and other Catalan language models.

Sources: Catalan Wikipedia (licensed under CC BY-SA 4.0).

Supported Tasks and Leaderboards

• —Extractive Question Answering
• —Evaluation for Catalan Language Models

Languages

• —Catalan (ca-ES)

Dataset Structure

Data Instances

{
  "id": "P_66_C_391_Q1",
  "title": "Xavier Miserachs i Ribalta",
  "context": "En aquesta època es va consolidar el concepte modern del reportatge fotogràfic...",
  "question": "De què es diferenciava el reportatge fotogràfic?",
  "answers": [{
    "text": "del fotoperiodisme[n. 2] i de la fotografia documental",
    "answer_start": 92
  }]
}

Data Fields

Following SQuAD v1.0:

• —id (str): Unique identifier.
• —title (str): Title of the Wikipedia article.
• —context (str): Text passage from Wikipedia.
• —question (str): Question posed about the context.
• —answers (list of dicts): Each answer contains:

• —text (str): Answer span from the context.
• —answer\_start (int): Character offset of the answer.

Data Splits

• —train: 11,259 examples
• —dev: 1,493 examples
• —test: 1,428 examples

Dataset Creation

Curation Rationale

Designed to provide a high-quality resource for Catalan, a low-resource language, to train and evaluate extractive QA systems.

Source Data

• —Collection: Articles scraped from the Catalan Wikipedia.
• —Selection: 597 high-quality, non-translated articles.
• —Contexts: 3,111 contexts (5–8 sentences each) extracted.
• —Questions: 1–5 questions created per context → 15,153 QA pairs.
• —Guidelines: Adapted from SQuAD 1.0.

Source Language Producers

Volunteer contributors to Catalan Wikipedia.

Annotations

• —Process: Professional annotators created questions and answers following adapted SQuAD guidelines.
• —Annotators: Commissioned to a specialized company with native Catalan speakers.

Personal and Sensitive Information

No personal or sensitive information is included.

Considerations for Using the Data

Social Impact

Supports the development of NLP for Catalan, reducing the resource gap for minority languages.

Discussion of Biases

• —May inherit coverage biases present in Wikipedia.
• —Articles selected are high-quality but do not fully represent all domains of Catalan usage.

Other Known Limitations

• —Limited domain (Wikipedia) may reduce generalizability.

Additional Information

Dataset Curators

• —Text Mining Unit (TeMU), Barcelona Supercomputing Center (BSC)

• —Contact: bsc-temu@bsc.es

Funding

Funded by the Departament de la Vicepresidència i de Polítiques Digitals i Territori de la Generalitat de Catalunya, within the framework of Projecte AINA.

Licensing Information

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Citation Information

@inproceedings{armengol-estape-etal-2021-multilingual,
    title = "Are Multilingual Models the Best Choice for Moderately Under-resourced Languages? {A} Comprehensive Assessment for {C}atalan",
    author = "Armengol-Estap{\'e}, Jordi  and
      Carrino, Casimiro Pio  and
      Rodriguez-Penagos, Carlos  and
      de Gibert Bonet, Ona  and
      Armentano-Oller, Carme  and
      Gonzalez-Agirre, Aitor  and
      Melero, Maite  and
      Villegas, Marta",
    booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.findings-acl.437",
    doi = "10.18653/v1/2021.findings-acl.437",
    pages = "4933--4946",
}

DOI: 10.5281/zenodo.4562344

Contributions

N/A