CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data

TLDR

Pre‑training text representations have substantially improved many NLP tasks, and their performance scales with corpus size provided the data quality remains high. The authors present an automatic pipeline that extracts large, high‑quality monolingual datasets from Common Crawl for multiple languages. The pipeline adopts fastText’s deduplication and language‑identification steps and adds a quality filter that selects documents resembling Wikipedia‑style corpora.

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.