Question Recommendation Mechanism In Community Question Answering Systems

With the fast-development of information network technology, the community question answering system as a typical application of Web2.0 has become an important channel for the modern network information acquisition and knowledge sharing. There are some very popular systems, for instance, Baidu Zhidao, Yahoo! Answers, Zhihu and so on. These systems make full use of the community users’ knowledge to meet the universal needs of information.From the current development situation, although the community question answering system has attracted a large number of users to participate, there are still many user questions without reply. There is also a considerable part of the answer that the users are not satisfied with, in the problem has been resolved. One of the reasons for these phenomena is that although there are a large number of users, even experienced users, it is difficult to find interested question in from a large number of new issue to solve problems.Therefore, this dissertation deeply studied the problems recommendation mechanism in community question answering system for helping find potential users can answer question and then effectively recommend question. It can make a lot of new questions get timely and accurate answers. Thus, this dissertation proposes a question recommendation algorithm of combining user interest and user activity information. The main research work is as follows:(1)Estimate user active. Analyze the important effect of community users’ activity information in the study of question recommendations, and make user’s behavior as a data stream, using a statistic method based on exponentially decaying window to estimate the user’s activity value.(2)Create user interest model. Making full use of the rich personalized information of community question answering system users to establish question and answering information file, and then modeling of user interest through the way of based on the theme. Thus, each user’s interest can be expressed by the underlying theme distribution.(3)Add the user active in the recommendation algorithm. Combined with the user interest and activity information for optimizing recommendation method, and get the users’ comprehensive score ranking result, so as to find the right answer of question without answer.(4)Create experiment system. In order to test the optimized effect of users’ active information characteristics on the sequencing results, this dissertation design and implement the experimental system to compare the experimental method. The experimental results show that the improved question recommendation algorithm improves the performance of the algorithm.