An Improved Word Vector-Based Symptom Extraction Method for Traditional Chinese Medical Record Analysis

Abstract

Extracting and standardizing symptoms from traditional Chinese medical records plays an important role in intelligent diagnosis. Recently, abundant word vector models have been developed and used in natural language processing tasks due to their powerful performance. However, simply using a word vector model as core to analysis text is hard to satisfy both time and precision requirements. To improve this situation, we introduce an improved word vector-based symptom extraction method for traditional Chinese medicine which can extract and standardize symptoms in original medical texts written in Chinese. We design this method into three parts, Word Segmentation, Word Vector Generation, and Term Substitution. Experimental results on our dataset show that our method has a good effect in extracting medical symptoms and discarding redundant words. Compared to other baseline models of word vector representation, our method performs well in general performance of efficiency and accuracy.