基于条件随机场的医药领域症状信息抽取

Information extraction of symptoms in the medical field based on CRF

为了实现中文医药领域症状信息的自动化抽取,提出了一种基于条件随机场的拆分症状文本特征的抽取方法,将症状文本自动识别拆分为症状主体和症状表现形式自动识别两个部分,再将这两种识别结果作为特征加入到症状信息抽取过程中。信息抽取的结果包含完整的症状信息二元组:症状主体和症状表现形式。实验表明,该方法在症状信息抽取的准确率及查全率上有较大提升。

In order to achieve the automated extraction of Chinese language symptom information, we put forward an extraction approach of splitting text features of symptoms based on CRF. The approach splits symptom tests into symptom subject recognition and symptom manifestation recognition, and then uses the results in the extraction of symptoms. The final results include complete two-tuples： symptom subject, symptom manifestation. Our experi- ments show that this approach can achieve a higher recognition rate than other methods.