Quality Does Matter: A Detailed Look at the Quality and Utility of Web-Mined Parallel Corpora
This work addresses the challenge of data quality for low-resource machine translation, showing incremental improvements in leveraging web-mined corpora effectively.
The study analyzed the quality of web-mined parallel corpora for low-resource language pairs, finding significant quality variations across datasets and languages, and demonstrated that NMT models trained on the top 25k portion of some web-mined datasets can match human-curated datasets in performance.
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.