Zero-Shot Cross-Lingual Ranking Performance on Synthetic Code-Switched Data Variations

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: How does the performance of zero-shot cross-lingual ranking models vary when trained on different types of synthetic code-switched data (e.g., word-level vs. phrase-level switching) as measured by MRR on multilingual benchmarks like XNLI or MLQA?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.7/10.