Improving Zero-Shot Cross-Lingual Retrieval with Multilingual Contrastive Learning

Transferring information retrieval (IR) models from a high-resource language (typically English) to other languages in a zero-shot fashion has become a widely adopted approach. In this work, we show that the effectiveness of zero-shot rankers diminishes when queries and documents are present in different languages. Motivated by this, we propose to train ranking models on artificially code-switched data instead, which we generate by utilizing bilingual lexicons. To this end, we experiment with lexicons induced from (1) cross-lingual word embeddings and (2) parallel Wikipedia page titles. We use

Research goal: Can the effectiveness of code-switched training data for zero-shot cross-lingual retrieval be further improved by incorporating multilingual contrastive learning objectives, as measured by performance on the MTop and FEVERous benchmarks?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.5/10.