Empowering Language Model with Guided Knowledge Fusion for Biomedical Document Re-ranking

Pre-trained language models (PLMs) have proven to be effective for document re-ranking task. However, they lack the ability to fully interpret the semantics of biomedical and healthcare queries and often rely on simple patterns for retrieving documents. To address this challenge, we propose an approach that integrates medical knowledge into PLMs to guide the model toward effectively capturing information from external sources and retrieving the correct documents. We performed comprehensive experiments on two biomedical datasets and an open-domain dataset. We demonstrate the capability of the proposed mutual information-based feature fusion technique by comparing it with the existing feature fusion techniques. Our extensive experiments on multiple datasets show that our proposed approach significantly improves vanilla PLMs and other existing approaches for document re-ranking task in the biomedical/clinical domain.