Research On Expert Users' Behavioral Characteristics In Community Question Answering

A huge number of web pages are on the internet but the quality of these pages is uneven.Users could not quickly find all the answers to their problems with search engines.The Community Question Answering(CQA)supplement the search engine.From the first fee-based service to all free services,CQA provides a platform for users to ask and answer so that problems can be solved quickly and users can share and get knowledge as well as ideas.The quality of CQA depends on expert users who provide the main content of CQA.Therefore,expert users are very important to CQA.The study on expert users' behavioral characteristics can show the distribution of expert user behavior characteristics,explore the correlations between the behaviors and achieve the division of expert users.These findings would play an important role in finding,recommending and maintaining expert users in CQA.The paper makes a research on expert users' behavioral characteristics in CQA by combining the researches on expert user finding in CQA and user behavior characteristics in virtual community.The paper chooses Zhihu as the data source and captures the question data,answer data and user data in four different topics by crawling the pages of Zhihu.Then,the weighted PageRank algorithm which has been proved to be effective is used to calculate the expert value of users.The paper defines top 10%of users as expert users according to the existing researches.Moreover,the paper compares these expert users with expert users found by other two methods based on features to testify that the method used in the paper is effective.The statistical analysis,correlation analysis,cluster analysis and social network analysis are used to analyze expert users' behavioral characteristics.The behaviors of answering,asking and following are included.The answering behavioral characteristics consist of answering number,average responding time,answer activeness and the vote for answers.The number of questions is the main characteristic of asking behavior.The characteristic of following behavior is measured by the number of topics followed by users.K-means++ is used to cluster expert users.The paper optimizes the number of clusters and divides the expert users into five types.Social network analysis is used to get the density,clustering coefficient and components of expert users' networks based on answering the same questions in different topics and different stages.According to the analysis of expert user behavior,the paper finds the numerical distributions of the characteristics of the expert users' answering,asking and following behaviors and the correlations between different behavioral characteristics.According to the clustering analysis,expert users can be divided into 5 types.Finally,some suggestions based on the findings are proposed to promote CQA.Through the paper,the research on user behavior analysis in CQA is refined and expanded and the results can be applied to make CQA more perfect.However the selection of expert users is not strictly precise and the in-depth analysis is needed.In the future research,the expert users' selection in different topics could be optimized,the textual characteristics could be analyzed and other social network networks can be used to make a comparison.