Answer Key Word Extraction And Quality Evaluation Based On Opening Questions Of CQA

With the increasing popularity of Internet,Community-Based Question Answering(CQA)has become an important platform for users' information accessing and knowledge sharing.However,such quality problems as redundancy,unreliability,incompleteness and others still exist in the answering resources in CQA,which has impeded the sharing and reusing of answering resources.Under this situation,a lot of research has been done on summary of answers and evaluation of the quality of answers in recent years,yet the research on open questions with more than one answer is still insufficient.By studying on the obtaining of key words of answers and evaluation of the quality of answers,this paper addressed the problems including one-sidedness and incompleteness that existed in the answers to open questions,thereby improving the quality of answers acquiring and enhancing the efficiency of people's information obtaining.This paper proposed a multi-feature and multi-disciplinary algorithm for extraction of keywords in answers,which integrated LDA topic models,semantic resources,statistical characteristics and contextual features.Seed keywords in a certain field are firstly obtained by topic model and then expanded in two ways:context and synonyms.Compared with traditional methods for keyword acquisition and expansion,this algorithm is able to provide keywords with larger scale and higher accuracy to some extent.This paper also put forward a semantic-based algorithm for evaluation of the quality of answers.First,the keywords of corresponding field are used to filter the answers;then a CNN(convolutional neural network)model is used to train the quality evaluation model.Finally,the quality of the answers can be tested by the model.Experimental result of the data set from “Zhihu”(the Chinese Quora)demonstrated that,compared with traditional methods for evaluation of quality of answers,the algorithm provided in this paper clearly presented higher accuracy.