README.md

metadata

annotations_creators:
  - no-annotation
language:
  - pt
language_creators:
  - found
license:
  - mit
multilinguality:
  - monolingual
pretty_name: brwac
size_categories:
  - 10M<n<100M
source_datasets:
  - original
tags:
  - ufrgs
  - nlp
  - brazil
task_categories:
  - fill-mask
task_ids:
  - masked-language-modeling

Dataset Card for BrWac

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Source Data
• Additional Information
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Homepage: BrWaC homepage
• Repository: BrWaC repository
• Paper: The brWaC Corpus: A New Open Resource for Brazilian Portuguese
• Point of Contact: Jorge A. Wagner Filho

Dataset Summary

The BrWaC (Brazilian Portuguese Web as Corpus) is a large corpus constructed following the Wacky framework, which was made public for research purposes. The current corpus version, released in January 2017, is composed by 3.53 million documents, 2.68 billion tokens and 5.79 million types. Please note that this resource is available solely for academic research purposes, and you agreed not to use it for any commercial applications. Manually download at https://www.inf.ufrgs.br/pln/wiki/index.php?title=BrWaC

This is a Tiny version of the entire dataset for educational purposes. Please, refer to https://github.com/the-good-fellas/xlm-roberta-pt-br

Supported Tasks and Leaderboards

Initially meant for fill-mask task.

Languages

Brazilian Portuguese

Dataset Structure

Data Instances

{
  "docid": "netg-19eac6", 
  "title": "Pra\u00c3\u00a7a Eugenio Jardim, Rio de Janeiro (Capital) - Trovit Im\u00c3\u00b3veis", 
  "uri": "http://rd.trovit.com/cookieid.1435413615788084558a649.85861732/id.1B-1S1lw1zBB/what_d.casa%20copacabana/origin.1/section.1/section_type.1/pos.10/tracking/page.1/country.br/vertical.homes/iuid.23286http-36578872c35d9d6/type.5/", 
  "sentences": [
    "Pra\u00e7a Eugenio Jardim Rio de Janeiro Capital", 
    "Pra\u00e7a Eugenio Jardim Rio de Janeiro Rio de Janeiro", "R$1.400", 
    "Descri\u00e7\u00e3o", "\u00d3timo apartamento em Copacabana no corte do Cantagalo bem em frente ao metr\u00f4 pr\u00f3ximo lagoa praia", 
    "Muito bem localizado uma \u00e1rea nobre", 
    "apartamento bem espa\u00e7oso tem quartos uma depend\u00eancia cozinha \u00e1rea de servi\u00e7o tr\u00eas banheiros sala", 
    "Tem mais de um quarto dispon\u00edvel porque gente ia entregar apartamento esse quarto de casal divide banheiro com apenas outro quarto", 
    "Seremos pessoas em casa no total", 
    "Este quarto j\u00e1 est\u00e1 dispon\u00edvel outro quarto de casal estar\u00e1 dia 10 04", 
    "Estamos procurando algu\u00e9m pra dividir apartamento n\u00e3o sou propriet\u00e1ria n\u00e3o republica pra dividir casa mesmo sem regras s\u00f3 bom senso"
    ]
}

The given format differs from the original file. We choose to represent the text in full sentences as it is suitable for tasks like fill-mask.

Data Fields

• Docid
• Title
• Uri
• Sentences

Data Splits

• Train
• Validation
• Test

Dataset Creation

Personal and Sensitive Information

All data were extracted from public sites.

Licensing Information

MIT

Citation Information

@inproceedings{wagner2018brwac,
  title={The brwac corpus: A new open resource for brazilian portuguese},
  author={Wagner Filho, Jorge A and Wilkens, Rodrigo and Idiart, Marco and Villavicencio, Aline},
  booktitle={Proceedings of the Eleventh International Conference on Language Resources and Evaluation 
  (LREC 2018)},
  year={2018}
}

Contributions

Thanks to @the-good-fellas for adding this dataset as hf format.