The Research And Implementaltion Of Expert Finding Method For Community Question Answering

With the development of web 2.0,popularity of website services which enhanced communications is continuously increasing.Among these services,the interactive question answering community(Baidu Zhidao and Stack Overflow)provides an onsite platform for users to propose questions and give answers.The emergence of these web sites help net users to achieve the transmission of information and knowledge sharing.But with the continuous growth of users in community question answering(CQA),a lot of problems come up.For example,many proposed questions don't get timely answer and for the problem that has been resolved,there is a big part of the answers that can not satisfied the questioner.Therefore,it's of great practical significance to find expert users from the massive user in CQA sites.In recent years,with the continuous development of CQA sites,related research about it has found a great increase and has made great breakthrough.Many scholars have proposed many different methods to solve the expert finding problem.The key idea behind these researches is mainly based on network structure and topic model.This paper focus on two different application scenes,question recommending and answer ranking respectively,and introduce two different expert finding methods based on topic model and link analysis.At the same time,for the problem about cold start user in question recommending,the paper introduce a cross-network expert finding method based on the common user.The main work is as follows:1)We study the user network structure in CQA.Users in CQA is linked by questions and answers and thus forming a unique user relational network.By analyzing the network structure of CQA,an effective construction of user network can be introduced.2)We introduce the expert finding method based on user interest model.Focused on the scene about question recommending,we introduced an expert finding method based on the classical random walk method,which can be well combined with topic model.The experiment results validate the efficiency of proposed method.3)We introduce the expert finding method based on link analysis.Focused on the scene about answer ranking,we introduced an expert finding method based on the link analysis and topic model.The experiment results validate the efficiency of proposed method.4)We introduce the cross-network expert finding method based on the common user.Focused on the scene about cold-start user recommendation,we introduced an expert finding method based on the common user to overcome the problem of insufficient user information.The experiment results validate the efficiency of proposed method.