A NLP Application of Automated Symptom Information Extraction from TCM Medical Cases

Abstract

Starting from natural language processing (NLP) technology, we compare the extraction results of TFIDF and Word2vec methods to explore research ideas that are more applicable to automated extraction of symptom information of traditional Chinese medicine (TCM) medical cases, and provide reference for the development of automated analysis of TCM medical cases. On the basis of constructed medical case dictionary, TFIDF and Word2vec methods were used to extract symptoms from heart cases, and the results were compared and analyzed. In medical cases, the onset of patients was often accompanied by palpitations, chest tightness, chest pain, shortness of breath, dizziness and other symptoms, and certain associations between symptoms were also found. The results of experimental evaluation showed that accuracy and recall rates of Word2vec method extraction were higher than those of TFIDF method. Compared with TFIDF method, Word2vec method is more effective when applied to the task of automated symptom information extraction from TCM cases.