Design And Research Of Intelligent Question Answering System Based On Chinese Community

With the fast development and popularization of Web2.0 and mobile Internet,the Online Community Question and Answer(QA)Platform has gradually become the main channel for people to exchange information and consult.In the past few years,major Online Community QA Platforms have accumulated a large deal of information about QA,which lead to the coming out of the new high-end AI services with intelligent QA.However,in the process of general nature language question and answer,there are some problems such as logic being described complexly,sentence structure being not neat and providing incorrect or repeated answers,which make it difficult for information retrieval and machine learning algorithms to provide accurate and effective answers.Aiming at the noise existing in the Integrated Community Intelligent QA Platform,this thesis proposes an optimized method focusing on the database of community knowledge.The new method sorts several answers gathered in one community answer set by analyzing the public non-text attribute,e.g.professional level,field of expertise,satisfaction degree of satisfaction,which eliminates the influence brought by the ‘Multi-answer questions' phenomenon.Aiming at the problem that the QA system retrieves answers inaccurately,an optimized retrieval sequencing method based on reliability evaluation mechanism is proposed.The new method firstly uses the matched text to search out the candidate answers.Then,the content searched form the new question of the user is hierarchically classified and matched with the fields that the responders are good at.Finally,it sorts the candidate answers again according to the trustworthy weight obtained from the matched result,which improves the hit rate of the QA system to the data of the community platform.The solutions proposed in this thesis is applied to Chinese Online Health Community data and a QA system for assisting diagnosis and treatment is established based on the data.Experimental results show that,the performance and the accuracy of the optimized system are improved explicitly.