Research On Personalized Service Based On Question Answering Community

Following the rapid development of internet, there has been an increasing number of information online. It is a very hot research direction in the filed of how to effectively find the needed information in the mass data. Although now there are many search engines which can help people search information their needed, they also have some disadvantages, such as the method of keyword searching often can't find exact results. With the advent of information society and knowledge economy, this traditional passive impartment of knowledge has given place to the initiative research of knowledge. People not only settle for browsing and keyword searching, but also hope to express their inclination and questions using natural language processing (NLP).In recent years, new generation search engines- question answering communities (QAC) have become a burgeoning mode of knowledge sharing. These communities don't inquire the contents online directly based on keywords, but submit questions based on specific needs by users themselves. They also encourage other users to create the answers to the questions by incentive mechanism. Meanwhile, these answers to the questions are further taken as searching results and provide them to other users with similar questions. People are not only the users of search engines but also creators of knowledge [1]. In this paper, we research deeply on personalized services of question answering communities combining the technologies of data mining with web searching. The key researches are as follows:(1) Results clustering algorithms based on QA communities services. According to known text clustering algorithms and applications in the filed of search engine, we propose a results clustering algorithms based on QAC. By analyzing the QA pairs returned by QA communities, we acquire keywords in the results using our algorithm and take them as candidate labels, users can get right QA results which accord with their knowledge need by choosing some labels.(2) QAC experts recommending algorithm based on weighted HITS algorithm. Since QAC users hope the personalization needs of experts'help, we analyze the different roles which every user act in the QAC, and define different weights according to users'interactive actions. We calculate user's score based on weighted HITS algorithm and recommend experts to the users who want to get help from our QAC.