GlotCC: An Open Broad-Coverage CommonCrawl Corpus and Pipeline for Minority Languages

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kargaran, Amir Hossein, Yvon, François, Schütze, Hinrich
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909523436896256
author Kargaran, Amir Hossein
Yvon, François
Schütze, Hinrich
author_facet Kargaran, Amir Hossein
Yvon, François
Schütze, Hinrich
contents The need for large text corpora has increased with the advent of pretrained language models and, in particular, the discovery of scaling laws for these models. Most available corpora have sufficient data only for languages with large dominant communities. However, there is no corpus available that (i) covers a wide range of minority languages; (ii) is generated by an open-source reproducible pipeline; and (iii) is rigorously cleaned from noise, making it trustworthy to use. We present GlotCC, a clean, document-level, 2TB general domain corpus derived from CommonCrawl, covering more than 1000 languages. We make GlotCC and the system used to generate it - including the pipeline, language identification model, and filters - available to the research community. Corpus v. 1.0 https://huggingface.co/datasets/cis-lmu/GlotCC-v1, Pipeline v. 3.0 https://github.com/cisnlp/GlotCC.
format Preprint
id arxiv_https___arxiv_org_abs_2410_23825
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle GlotCC: An Open Broad-Coverage CommonCrawl Corpus and Pipeline for Minority Languages
Kargaran, Amir Hossein
Yvon, François
Schütze, Hinrich
Computation and Language
Artificial Intelligence
The need for large text corpora has increased with the advent of pretrained language models and, in particular, the discovery of scaling laws for these models. Most available corpora have sufficient data only for languages with large dominant communities. However, there is no corpus available that (i) covers a wide range of minority languages; (ii) is generated by an open-source reproducible pipeline; and (iii) is rigorously cleaned from noise, making it trustworthy to use. We present GlotCC, a clean, document-level, 2TB general domain corpus derived from CommonCrawl, covering more than 1000 languages. We make GlotCC and the system used to generate it - including the pipeline, language identification model, and filters - available to the research community. Corpus v. 1.0 https://huggingface.co/datasets/cis-lmu/GlotCC-v1, Pipeline v. 3.0 https://github.com/cisnlp/GlotCC.
title GlotCC: An Open Broad-Coverage CommonCrawl Corpus and Pipeline for Minority Languages
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2410.23825