Multi-Field Hybrid Retrieval-Augmented Generation for Maritime Accident Root Cause Analysis
Researchers have developed a novel retrieval-augmented generation (RAG) framework designed to automate the root cause analysis (RCA) of maritime accidents. This system utilizes a structured knowledge base derived from over 13,000 Korea Maritime Safety Tribunal (KMST) reports, indexing key fields like summary, causes, and disposition. The framework employs a field-aware hybrid retrieval strategy, combining sparse and dense rankings, which significantly outperforms baseline methods. By grounding a generative model on these retrieved precedents, the system enhances the quality of RCA drafting, making investigations more efficient and consistent. AI
IMPACT This framework could significantly streamline maritime safety investigations by automating precedent search and improving the consistency of accident reports.