CrossRank

CrossRank is a cross-lingual information retrieval system that enables English queries to retrieve relevant Japanese documents using multilingual embeddings and vector search.

User feedback is applied using the Rocchio algorithm to refine the query vector and improve rankings. The system is evaluated using MRR@10 and Recall@100, showing clear improvements after feedback.

Note: This project was developed as an exploration of cross-lingual retrieval and relevance feedback systems. The evaluation uses benchmark queries and qrels from the Mr. TyDi Japanese dataset, with additional experiments conducted to study retrieval behavior under different Rocchio parameter configurations.