Negation Detection in Chinese Electronic Medical Record Based on Rules and Word Co-occurrence
In order to extract negative terminologies in Chinese Electronic Record. Many methods have been developed. One popular and simple method is based on rules. However, the negative predictive value drops significant if the sentence contains several kinds of punctuation. In our research, a new method is used to solve the problem. The new method combines rules with word co-occurrence. In the experiments, 200 medical texts including 150,865 Chinese characters are used to test the new method. The negative predictive value is 99.85 %, which is 7.85 % higher than the rule-based method. That is to say, this method can tolerate various kinds of punctuations existing in the sentences. Therefore, the value of false-positive probability drops obviously.
KeywordsWord co-occurrence Mutual information Negation detection
This work is supported by Nantong social undertakings technology innovation and demonstration program (No. HS2012045) and Natural Science Foundation of Nantong University (No. 11Z010) and Jiangsu Province modern technology education project (No. 2013-R-24890).
- 1.Haomin Li, Ying Li, Huilon Duan et al (2008) Term extraction and negation detection method in chinese clinical document. Chin J Biomed Eng 27:716–734Google Scholar
- 2.Yinghong Liang, Wenjing Zhang, Youcheng Zhang (2010) Term recognition based on integration of c-value and mutual information. Comput Appl Soft 27:108–110Google Scholar