Research On The Method Of Expert Contribution Evaluationin Knowledge Community

With the rapid development of Web2.0, the interaction between experts and users of Knowledge Community(KC) becomes a popular trend of the network. In order to provide individualized services for users in KC, Expert contribution evaluation(ECE) is a necessary part of research in the community for the prosperity and development of KC. In order to offer a bridge of communication between experts and users, KC must have numerous experts and a rich knowledge database and provide quickly and timely responses to newly submitted questions of users. In all studies of KC referenced in this paper, submitted questions by users will be answered by experts who are good at professional field knowledge. Newly submitted question by users can have promptly and accurately answer and the knowledge database will be expanded then if among huge experts we find out experts which contributions to KC and expert authority are very high. So, Expert evaluation in KC is very important and far-reaching and it is an effective way to development KC rapidly. How to identify experts with high quality from massive experts in KC has become a study of KC in an extremely important work.In this paper, we introduce research background and existing research results of ECE in KC. We also studies KC and problems of evaluation in detail. At present, Link analysis method of Google’s core technology and LDA topic model method are widely used in ECE. This paper is aimed at study problems of ECE in CK based on these two kinds of methods. The following is the main work of this paper:Firstly, in this paper, we analyze question answering link structure between experts and users in KC, and then construct link relation direct graph between experts and users. On one hand, this method uses LDA topic model to find topic distribution of question answering which belongs to professional field and experts which are involved in professional field, on the other hand, we present Knowledge Community Expert Professional Field(KCEPF) topic mining model which is suitable for KC.Secondly, we present a method to evaluate contribution of expert based on Professional Field Penetration(PFP), which uses KL distance to calculate the similarity between professional fields and experts’ penetration in various professional fields. The contribution in KC of expert is a linear combination of the contribution in this professional field and the contribution in other similar professional fields.Thirdly, according to the problem of uniform transfer expert contribution value in Page Rank Algorithm, the inadequate of traditional PageRank algorithm only depends on the link relationship between experts and users to rank. Based on combining with expert own behavioral characteristics in KC, we present three evaluation indicators which include the activity of expert, the popularity of expert and the knowledge contribution ability of expert, then add them to make some improvements to the traditional PageRank algorithm.The data sets used in the experiments are collected from Baidu Know. To evaluate the performance of the methods, we compare our methods with several expert authority calculation methods. To evaluate the performance for ECE, we used two widely-used metrics AP@10 and MAP. In professional field penetration-based method the expert’s contribution in similar professional field can improve the expert’s contribution in question answering which belongs to professional field. And the improved PageRank algorithm is more accurate in rank. The results show that two kinds of the method given in this paper to improve the performance of ECE in KC.