Bitext Mining for Low-Resource Languages via Contrastive Learning

Code

• github.com/steventan0110/align-filter
  OfficialIn paperpytorch★ 5

Abstract

Mining high-quality bitexts for low-resource languages is challenging. This paper shows that sentence representation of language models fine-tuned with multiple negatives ranking loss, a contrastive objective, helps retrieve clean bitexts. Experiments show that parallel data mined from our approach substantially outperform the previous state-of-the-art method on low resource languages Khmer and Pashto.

Tasks

Reproductions