CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data

Guillaume Wenzek; Marie-Anne Lachaux; Alexis Conneau; Vishrav Chaudhary; Francisco Guzmán; Armand Joulin; Edouard Grave

CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data

Computation and Language 2019-11-18 v2 Information Retrieval Machine Learning Machine Learning

Authors: Guillaume Wenzek , Marie-Anne Lachaux , Alexis Conneau , Vishrav Chaudhary , Francisco Guzmán , Armand Joulin , Edouard Grave

View on arXiv ↗ PDF ↗

Abstract

Pre-training text representations have led to significant improvements in many areas of natural language processing. The quality of these models benefits greatly from the size of the pretraining corpora as long as its quality is preserved. In this paper, we describe an automatic pipeline to extract massive high-quality monolingual datasets from Common Crawl for a variety of languages. Our pipeline follows the data processing introduced in fastText (Mikolov et al., 2017; Grave et al., 2018), that deduplicates documents and identifies their language. We augment this pipeline with a filtering step to select documents that are close to high quality corpora like Wikipedia.

Keywords

information extraction pipeline optimization corpus

Cite

@article{arxiv.1911.00359,
  title  = {CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data},
  author = {Guillaume Wenzek and Marie-Anne Lachaux and Alexis Conneau and Vishrav Chaudhary and Francisco Guzmán and Armand Joulin and Edouard Grave},
  journal= {arXiv preprint arXiv:1911.00359},
  year   = {2019}
}

CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data

Abstract

Keywords

Cite

Related papers