TY - GEN
T1 - Reliable Typical Case Diagnosis via Optimized Retrieval-Augmented Generation Techniques
AU - Zhang, Kaiyuan
AU - Wang, Bo
AU - Yuan, Changsen
AU - Feng, Chong
AU - Shi, Ge
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.
PY - 2025
Y1 - 2025
N2 - Recent advances in Large Language Models (LLMs) have shown remarkable capabilities in various Natural Language Processing (NLP) tasks. However, their performance in specialized domains, particularly healthcare, is still limited by the lack of domain expertise and output reliability. In this paper, we present a retrieval-augmented framework for the CHIP 2024 Diagnostic Consistency Task in Typical Medical Case Records. Our approach addresses these limitations by combining the generative power of LLMs with the retrieval of relevant medical knowledge. Specifically, we introduce a hybrid retrieval mechanism that integrates traditional BM25 with dense retrieval methods, along with a context compression strategy and self-consistency verification module. This framework enables more reliable and accurate diagnostic predictions by leveraging both contextual similarities and domain-specific knowledge. Our method achieves an average F1 score of 0.9517 on the test set, demonstrating the effectiveness of our approach for specialized medical diagnosis tasks.
AB - Recent advances in Large Language Models (LLMs) have shown remarkable capabilities in various Natural Language Processing (NLP) tasks. However, their performance in specialized domains, particularly healthcare, is still limited by the lack of domain expertise and output reliability. In this paper, we present a retrieval-augmented framework for the CHIP 2024 Diagnostic Consistency Task in Typical Medical Case Records. Our approach addresses these limitations by combining the generative power of LLMs with the retrieval of relevant medical knowledge. Specifically, we introduce a hybrid retrieval mechanism that integrates traditional BM25 with dense retrieval methods, along with a context compression strategy and self-consistency verification module. This framework enables more reliable and accurate diagnostic predictions by leveraging both contextual similarities and domain-specific knowledge. Our method achieves an average F1 score of 0.9517 on the test set, demonstrating the effectiveness of our approach for specialized medical diagnosis tasks.
KW - Case Diagnosis
KW - Large Language Model
KW - Retrieval-Augmented Generation
UR - http://www.scopus.com/inward/record.url?scp=105003857060&partnerID=8YFLogxK
U2 - 10.1007/978-981-96-4298-4_19
DO - 10.1007/978-981-96-4298-4_19
M3 - Conference contribution
AN - SCOPUS:105003857060
SN - 9789819642977
T3 - Communications in Computer and Information Science
SP - 214
EP - 225
BT - Health Information Processing. Evaluation Track Papers - 10th China Health Information Processing Conference, CHIP 2024, Proceedings
A2 - Zhang, Yanchun
A2 - Chen, Qingcai
A2 - Lin, Hongfei
A2 - Liu, Lei
A2 - Liao, Xiangwen
A2 - Tang, Buzhou
A2 - Hao, Tianyong
A2 - Huang, Zhengxing
A2 - Lei, Jianbo
A2 - Li, Zuofeng
A2 - Zong, Hui
PB - Springer Science and Business Media Deutschland GmbH
T2 - 10th China Health Information Processing Conference, CHIP 2024
Y2 - 15 November 2024 through 17 November 2024
ER -