Research On The Evaluation Of Answer Quality In Q&A Communities Based On Multiple Models

The rapid development of information technology provides a new tool for people to acquire and share knowledge.The representative Q&A community plays an important role in people's study,life,work,entertainment and other aspects.With the exponential growth of information volume,how to effectively evaluate the quality of answers in the Q&A community has become an important issue.work.In order to evaluate the quality of the answers generated by users in the Q&A community,this paper proposes a semantic feature and non-semantic features that can be used as an answer quality evaluation in the deep mining question and answer community based on multiple models.The machine learning classification method is used to evaluate the quality of the answers.It mainly includes the following contents:(1)Quality evaluation of answers based on topic relevanceThe question and answer texts of the question and answer community have a textual feature matrix sparse,the context information is lost,and the length of the question text and the length of the answer text are seriously mismatched.This paper proposes a semantic extension based on the problem text.Answer quality evaluation model.Firstly,according to the text set composed of the problem and the correct answer,the Word2Vec training word vector is used to mine the text semantic information on the word granularity,and the main word of the problem text is calculated and its related word set,thus realizing the semantic expansion of the problem text..Second,using the LDA topic mining model,the text representation is the subject vector by inferring the subject of the question and answer text,and the subject relevance of the question and answer text is calculated using the JS distance.Finally,based on the topic relevance of questions and answers,the machine quality learning algorithms SVM,LR and decision tree C4.5 are used to construct the answer quality evaluation classifier.Based on the experiments,it is found that the topic relevance based on the semantic extension method proposed in this paper is significantly higher than the LDA-based topic correlation calculation method,but the accuracy of the answer classification.It is lower than the indirect evaluation method of existing research.(2)Quality evaluation of answers with multiple featuresThis paper combines text features,user features,social interaction features and time series features,and deeply explores the data in the Q&A community.It proposes to use AHP to calculate user authority,and explores the emotion of the answer text through naive Bayesian algorithm.Polarity,at the same time,the classification model that extracts the answer timeliness,the richness of the answers and the number of answers,and combines the above-calculated problem answer topic relevance training answer quality.Implement SVM,LR,and decision tree C4.5 answer quality classifiers.Through experiments,it is found that the classification effect of the model on the quality of the answer is significantly higher than that based on the relevance of the subject of the question answer.