Research On Answer Summarization Of Opinion Questionsin The Q&A Community

Recently,online question and answer community has received much attention.Q&A communities such as Zhihu,Baidu Pai,Quora and Yahoo! Answers have accumulated many resources,including users and answer resources.However,there are two problems with such resources: on the one hand,the quality of answers is uneven and the content of answers is redundant;on the other hand,from the point of view of the users,users need to spend a lot of time and energy obtaining valuable,meaningful textual information inefficiently.This paper proposes a method of an answer summary of opinion-type questions generated in Q&A communities.The main contents are as follows:(1)Calculation of answer sentence similarity.First of all,text mining techniques are used to pre-process texts.Due to the new Internet words that are not in the current semantic and sentiment dictionary,the semantic similarity and sentiment value of the new words are calculated by using the label propagation algorithm.Then based on the idea of cloud model,this paper uses the cloud features and cloud similarity to calculate the semantic and sentiment similarities between the answer sentences.It is proved that the method of text similarity in this paper is more accurate than statistical and dictionary methods.(2)Generation of answers' summarization.Because the emotion of opinion answers is quite sharp,this paper uses the sentiment similarity that is calculated before to generate three categories of positive,negative and neutral emotion sentence clusters.Each cluster meeting the condition needs generates a sentimental summary of the answers.The improved PageRank algorithm is used to calculate the weight of the answer sentences and obtain the abstract sentences.Then a similarity ranking method based on the best answer template is proposed to order candidate abstract sentences.Finally,the method proposed in this paper is applied in the example,which is closer to the best summary generated by human,and especially has more advantages than other methods in reading readability and logic.This paper generates answer summarization of community opinion questions by the calculation of sentence similarity.Summarizations condense opinions and contents of large numbers of texts,offers better information service to community users and has practical values.