BioASQ team enhances biomedical QA with novel reranking and recovery strategies

Researchers from the DS@GT ARC BioASQ team have developed a novel pipeline for biomedical question answering, enhancing performance on the BioASQ Task 14b. Their system integrates multi-source query expansion, a fine-tuned MiniLM-based semantic reranking model, and Reciprocal Rank Fusion (RRF) for improved document ranking. A key innovation is a conditional weak-question recovery strategy designed to address queries with poor initial retrieval, employing semantic expansion and relationship-aware augmentation. This approach, along with post-retrieval pruning and output validation, significantly boosts retrieval robustness and MAP@10 performance on challenging question sets. AI

IMPACT Introduces novel techniques for improving retrieval robustness and performance in specialized biomedical question-answering systems.

RANK_REASON The cluster contains an academic paper detailing a new methodology for biomedical question answering. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• BioASQ Task 14b
• DS@GT ARC BioASQ team
• MiniLM
• OpenBioLLM
• Reciprocal Rank Fusion

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →