According to the characteristics of traditional Chinese medicine clinical cases, in this paper, we propose an automatic generation algorithm of extraction patterns, GEPTCMA, which is based on bootstrapping. The algorithm includes two models: TCM-RPAM and TCMW-MODEL. TCM-RPAM is a method for the acquisition of bi relations and bi relation patterns; TCMW-MODEL is used for acquiring the sets of the domain keywords in the traditional Chinese medicine. Our experimental results show the precision/recall of data extraction using this system is as good as those from the templates based on structured information extraction. The domain experts in traditional Chinese medicine also give us an affirmation.