A Hybrid Approach for Biomedical Question Answering: Combining Sparse and Dense Retrieval with LLM-based Reranking
Abstract
Biomedical question-answering systems typically involve retrieving relevant documents and then reranking them based on their relevance to the query.Traditional sparse retrievers, like BM25 often fail to capture semantic relationships.While the dense retrievers address these limitations, they can also miss relevant documents due to short queries, vocabulary mismatch, and document specificity issues stemming from embeddings.To address these types of challenges, we propose a hybrid approach that consolidates the strengths of both sparse and dense retrieval methods, resulting in better performance.This ensemble approach generates a comprehensive list of candidate documents, which is then passed through a LLM-based reranking model named ColBert, a fine-tuned late interaction mechanism that works on document relevance to refine the ranking of the documents.We use the Flan-T5 answer generation model to produce a final answer to the query.The experiments were performed on the BioASQ dataset, which remarkably demonstrated the effectiveness of our approach and showcased its ability to improve retrieval performance.
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.