lhoestq's picture
lhoestq HF staff
add dataset_info in dataset metadata
b838742
metadata
annotations_creators:
  - no-annotation
language_creators:
  - expert-generated
language:
  - es
license:
  - cc-by-sa-4.0
multilinguality:
  - monolingual
size_categories:
  - 10M<n<100M
source_datasets:
  - original
task_categories:
  - other
  - text-generation
  - fill-mask
task_ids:
  - language-modeling
  - masked-language-modeling
paperswithcode_id: sbwce
pretty_name: Spanish Billion Word Corpus and Embeddings
dataset_info:
  features:
    - name: text
      dtype: string
  config_name: corpus
  splits:
    - name: train
      num_bytes: 8950895954
      num_examples: 46925295
  download_size: 2024166993
  dataset_size: 8950895954

Dataset Card for Spanish Billion Words

Table of Contents

Dataset Description

Dataset Summary

The Spanish Billion Words Corpus is an unannotated Spanish corpus of nearly 1.5 billion words, compiled from different resources from the web. This resources include the spanish portions of SenSem, the Ancora Corpus, some OPUS Project Corpora and the Europarl, the Tibidabo Treebank, the IULA Spanish LSP Treebank, and dumps from the Spanish Wikipedia, Wikisource and Wikibooks.

This corpus is a compilation of 100 text files. Each line of these files represents one of the 50 million sentences from the corpus.

Supported Tasks and Leaderboards

This dataset can be used for language modelling and for pretraining language models.

Languages

The text in this dataset is in Spanish, BCP-47 code: 'es'.

Dataset Structure

Data Instances

Each example in this dataset is a sentence in Spanish:

{'text': 'Yo me coloqué en un asiento próximo a una ventana cogí un libro de una mesa y empecé a leer'}

Data Fields

  • text: a sentence in Spanish

Data Splits

The dataset is not split.

Dataset Creation

Curation Rationale

The Spanish Billion Words Corpus was created to train word embeddings using the word2vect algorithm provided by the gensim package.

Source Data

Initial Data Collection and Normalization

The corpus was created compiling the following resources:

All the annotated corpora (like Ancora, SenSem and Tibidabo) were untagged and the parallel corpora (most coming from the OPUS Project) was preprocessed to obtain only the Spanish portions of it.

Once the whole corpus was unannotated, all non-alphanumeric characters were replaced with whitespaces, all numbers with the token “DIGITO” and all the multiple whitespaces with only one whitespace.

The capitalization of the words remained unchanged.

Who are the source language producers?

The data was compiled and processed by Cristian Cardellino.

Annotations

The dataset is unannotated.

Annotation process

[N/A]

Who are the annotators?

[N/A]

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

The data was collected and processed by Cristian Cardellino.

Licensing Information

The dataset is licensed under a Creative Commons Attribution-ShareAlike 4.0 International license (CC BY-SA 4.0)

Citation Information

@misc{cardellinoSBWCE,
     author = {Cardellino, Cristian},
     title = {Spanish {B}illion {W}ords {C}orpus and {E}mbeddings},
     url = {https://crscardellino.github.io/SBWCE/},
     month = {August},
     year = {2019}
}

Contributions

Thanks to @mariagrandury for adding this dataset.