A Parsing Approach For Event And Relation Extraction Of Clinical Guideline

As the important reference data of clinical diagnosis and treatment, clinicalguideline can help physicians make the appropriate decisions for specific clinicalproblems. However, along with the development of clinical medicine, clinical guidelineis huge in number, artificial management is very difficult. Therefore, establishing a set ofguideline event data has an important significance for improving the level of clinicaldiagnosis and treatment.This thesis focusing on event and relation extraction of clinical guideline. And italso proposes a paring based approach for event extraction and event relation extractionof clinical guidline. Firstly, it analized and summed up the event model and eventrelation model of medication according to the clinical guideline corpus,; then it proceedthe lexical analysis and parsing of clinical guideline to extract the candidate event triggerwords. And matching the event model of clinical guideline based on event trigger wordsas kernel according to the syntactic and semantic features, so that it can identify events ofclinical guideline; In order to accurately extract the internal relationship between clinicalguideline events, it further extracted the event relations based on the semantic relationsbetween events and the event relation models, after the extraction of clinical guidlineevent; Finally, it optimized the extracted information of clinical guidline event to get thenet result.This thesis also assessed the method by the specific data of clinical guideline. Itachieves a better precision and recall rate, and proves the feasibility and effectiveness ofthe method propesed in this thesis.