基于规则和词共现的中文电子病历否定检出

摘要

对于中文电子病历文本中的否定术语的检出，目前有很多方法，基于规则的否定检出是比较常用的一种算法。但是该方法无法解决由于标点录入错误造成假阳性的问题。因此，在基于规则算法的基础上，提出一种基于词共现的否定检出算法，通过收集200份中文电子病历约150865个汉字字符进行实验，新方法的阴性预测值比基于规则的算法提高了7．85％。所以，基于规则和词共现的否定检出算法能够很好地降低由于标点录入错误而出现假阳性术语的概率。%There are many methods used in negative terms detection in Chinese electronic medical record at present,and rules-based nega-tion detection is a commonly used algorithm.However,it fails in solving false positive problem caused by punctuations inputting errors. Therefore,on the basis of rules-based algorithm,we propose a word co-occurrence based negation detection algorithm.The negative predictive value of new method improves 7.85% than the rules-based method in the experiment collecting 200 copies of Chinese medical records inclu-ding about 150 865 Chinese characters.Therefore,the rules and word co-occurrence based negation detection algorithms can well reduce the probability of false-positive terms occurred due to punctuations inputting errors.