We conducted a detailed analysis on the quality of web-mined corpora for two low-resource languages (making three language pairs, English-Sinhala, English-Tamil and Sinhala-Tamil). We ranked each corpus according to a similarity measure and carried out an intrinsic and extrinsic evaluation on different portions of this ranked corpus. We show that there are significant quality differences between different portions of web-mined corpora and that the quality varies across languages and datasets. We also show that, for some web-mined datasets, Neural Machine Translation (NMT) models trained with their highest-ranked 25k portion can be on par with human-curated datasets.
@article{arxiv.2402.07446,
title = {Quality Does Matter: A Detailed Look at the Quality and Utility of Web-Mined Parallel Corpora},
author = {Surangika Ranathunga and Nisansa de Silva and Menan Velayuthan and Aloka Fernando and Charitha Rathnayake},
journal= {arXiv preprint arXiv:2402.07446},
year = {2024}
}