A medical question and answer retrieval method based on entity inverted index and semantic vector fusion
By fusing entity inverted indexes with semantic vectors, the accuracy problem of biomedical retrieval systems when processing complex named entities is solved, achieving high-precision medical question-and-answer generation and improving the retrieval and generation quality of the system.
Patent Information
- Authority / Receiving Office
- CN Β· China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF SCI & TECH
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-09
AI Technical Summary
Existing biomedical retrieval systems neglect the essential relationships between entities when processing complex named entities, leading to inaccurate retrieval, ambiguity, and missing information, making it difficult to meet the requirements of accuracy and rigor in knowledge output in biomedical scenarios.
A medical question-answering retrieval method based on entity inverted index and semantic vector fusion is adopted. A standardized entity set is generated through entity linking and standardization. Combining entity overlap and semantic vector retrieval, the RRF algorithm is used for multi-way ranking integration to ensure the accuracy of retrieval results.
It significantly improves the retrieval accuracy and generation quality of question-answering systems in the biomedical field, corrects the ambiguity bias of semantic vectors when dealing with abbreviations, synonyms, or similar terms, and ensures the certainty and accuracy of retrieval results.
Smart Images
Figure CN122173632A_ABST