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.
|Title of host publication||Advances in Knowledge Discovery and Data Mining - 23rd Pacific-Asia Conference, PAKDD 2019, Proceedings|
|Editors||Zhi-Hua Zhou, Min-Ling Zhang, Qiang Yang, Sheng-Jun Huang, Zhiguo Gong|
|Number of pages||12|
|State||Published - 2019|
|Event||23rd Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2019 - Macau, China|
Duration: Apr 14 2019 → Apr 17 2019
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||23rd Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2019|
|Period||4/14/19 → 4/17/19|
Bibliographical noteFunding Information:
Acknowledgments. This work was supported by National Key R&D Program of China (No. 2017YFC0803700), NSFC grants (No. 61532021 and 61472141), Shanghai Knowledge Service Platform Project (No. ZF1213), and SHEITC.
© Springer Nature Switzerland AG 2019.
- Topic modeling
- Traditional Chinese medicine
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science (all)