Unsupervised Bilingual Lexicon Induction for Low Resource Languages

Many low-resource languages (LRLs) do not have bilingual lexicons, and due to the same reason, cannot benefit from the supervised Bilingual Lexicon Induction (BLI) techniques. To address this, unsupervised BLI (UBLI) techniques were introduced. A prominent technique in this line is structure-based UBLI. There have been numerous improvements to this core idea, however they have been experimented with independently of each other. In this paper, we investigate whether using these techniques simultaneously would lead to equal gains. We use the unsupervised version of VecMap, a commonly used structure-based UBLI framework, and carry out a comprehensive set of experiments using the LRL pairs, English-Sinhala, English-Tamil, and English-Punjabi. Our results show that using both static and contextual embeddings as well as applying linear transformation on them provide the best result, where it outperforms UVecMap with an average of 1.47 pr@1 across languages. We also release bilingual dictionaries for English-Sinhala and English-Punjabi.