Multimodal Model Performance in Zero-Shot Cross-Lingual Image-Text Retrieval with Code-Switched Training

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: To what extent do multimodal models (e.g., CLIP, BLIP) improve zero-shot cross-lingual retrieval performance on image-text retrieval tasks (e.g., COCO, Flickr30K) when trained on code-switched data versus monolingual data, as measured by R@1 and R@10 metrics?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.1/10.