Xia, FeiTian, Yuanhe2019-08-142019-08-142019Tian_washington_0250O_19692.pdfhttp://hdl.handle.net/1773/44342Thesis (Master's)--University of Washington, 2019In this study, we make a move to answer ranking task of medical community question answering (QA). The task of answer ranking has four different settings based on whether features from questions or other answers are used. We designed multiple approaches under each setting to explore how different features contribute to high answer quality. Experimental results on a Chinese Medical QA Dataset show although question-answer relevance is important, cross-answer features are more crucial to distinguish good answers from bad answers. Therefore, in order to become good consultants, it is first recommended that doctors pay their attention to write high quality answers. Finally, our case study demonstrates that good answers tend to show their concern to patient's feelings and to provide more tips for daily care.application/pdfen-USnoneAnswer RankingChineseMedicalQuestion AnsweringComputer scienceInformation scienceLinguisticsThesis