Research And Implementation Of A QA System By Answer Retrieval From QA Databases

About a decade has passed since entering the Web 2.0 era.With the quick development of the Internet and popularization of mobile terminals,Internet users are generating massive information,which results in the information overload problem.Search engine technologies need to improve accordingly to help users retrieve desired information in a fast and accurate manner.The keyword search has become relatively matured.However,in many cases,query results are unsatisfactory due to improper keyword selection,or inaccurate due to a search engine's inability to filter out massive web pages.Compared with the foregoing approach,a question answering(QA)system takes a natural language question as an input rather than keywords,and directly returns the most relevant answers.This kind of systems improves user satisfaction and lessens query time.Based on this background,the study of this thesis is about the QA system by answer retrieval from QA databases.The main research work includes the following three aspects.(1)By studying and analyzing the similarity algorithm of texts based on dependency relation,a new filtered and weighted dependency relation based similarity algorithm of texts is proposed.Experimental results show that the proposed algorithm achieves higher accuracy than baselines.(2)A new QA retrieval model fusing dependency relation based,Vector Space Model(VSM)based,and Convolutional Neural Network(CNN)based similarity algorithms is developed.The model learns the optimal weights for different similarity features by RankSVM.The experimental results show that the model has higher F1 value and accuracy than baseline methods.(3)Through the realization of the QA retrieval model,the QA system can be supported by using QA data set from Community Question Answering(CQA).The system is built by using Sina App Engine(SAE)and WeChat public number.