计算机科学
发病机制
中医药
嵌入
等价(形式语言)
人工智能
语义对等
自然语言处理
医学
语言学
替代医学
病理
语义网
哲学
语义计算
作者
Ying Zhang,Wendi Ji,Haofen Wang,Xiaoling Wang,Jin Chen
标识
DOI:10.1007/978-3-030-16148-4_38
摘要
Traditional Chinese medicine (TCM) is well-known for its unique theory and effective treatment for complicated diseases. In TCM theory, "pathogenesis" is the cause of patient's disease symptoms and is the basis for prescribing herbs. However, the essence of pathogenesis analysis is not well depicted by current researches. In this paper, we propose a novel topic model called Multi-Content embedding LDA (MC-eLDA), aiming to collaboratively capture the relationships of symptom-pathogenesis-herb triples, relationship between symptom-symptom, and relationship between herb-herb, which can be used in auxiliary diagnosis and treatment. By projecting discrete symptom words and herb words into two continuous semantic spaces respectively, the semantic equivalence can be encoded by exploiting the contiguity of their corresponding embeddings. Compared with previous models, topic coherence in each pathogenesis cluster can be promoted. Pathogenesis structures that previous topic modeling can not capture can be discovered by MC-eLDA. Then a herb prescription recommendation method is conducted based on MC-eLDA. Experimental results on two real-world TCM medical cases datasets demonstrate the effectiveness of the proposed model for analyzing pathogenesis as well as helping make diagnosis and treatment in clinical practice.
科研通智能强力驱动
Strongly Powered by AbleSci AI