Research on TCM Patent Annotation to Support Medicine R&D and Patent Acquisition Decision-Making

Traditional Chinese Medicine (TCM) patents contain abundant medical, economic and legal information. The effective analysis and mining of TCM patents is of great significance to support medicine R&D and patent acquisition decision-making. Named entity recognition is a key step in the research of TCM patents annotation. In order to solve the problem that the entity recognition in TCM patents relies on manual work in a large extent and the degree of automation is not high, in this paper, under the framework of ERNIE (Enhanced Language Representation with Informative Entities) pre-training language model, combined with the Bi-directional Gating Recurrent Unit (BiGRU) and Conditional Random Field (CRF), two important entities, herbal names and medicine effects, are recognized in TCM patent texts. Experimental results show that the performance of the method based on ERNIE-BiGRU-CRF is significantly improved compared with other models.