Using link topic model to analyze traditional Chinese Medicine Clinical symptom-herb regularities

Abstract

Traditional Chinese Medicine (TCM) is a clinical medicine, which focuses on human physiology, pathology, diagnosis and treatment of diseases. Numerous clinical practice and theory research in the TCM field have accumulated huge amount of data. These data include TCM basic databases, TCM literature, as well as a large number of databases or data warehouse on TCM clinical diagnoses and treatment. More and more people pay attention to the discovery of hidden regularities of TCM clinical data. In recent years, topic model has been popularly used for text analysis and information retrieval by extracting latent and significant topics from corpus. In this paper, we apply the Link Latent Dirichlet Allocation (LinkLDA), to automatically extract the latent topic structures which contain the information of both symptoms and their corresponding herbs. By experimental results, the latent topic with symptoms and their corresponding herbs show clinical meaningful results. Furthermore, the model is also compared with other topic models, such as author-topic model, and the result of LinkLDA got better results.