Research On Question Recommedation System Based On Weibo Community

This paper mainly focuses on the research about Q&A recommendation system based on microblog. Starting from the basic function of social microblogging and question answering system, the basic research is to realize the classification and recommendation of basic problems, algorithm classification and recommend research combining the characteristics of microblog. Combined with the characteristics of short text of question answering system and microblog, a classification algorithm for question is designed in this paper based on the problems of short text classification algorithm, using semantic information and other resources to enrich the representation of short text, improves the accuracy of short text similarity calculation, therefore improve the accuracy of the problem classification, the accuracy. At the same time, it is discussed that the algorithm of recommendation based on microblog user portrait and the ability of answering questions, problems of traditional recommendation algorithm are confined to the internal user interactive question answering system portrait, for a good question recommendation system, in terms of user scope the data is too narrow and small amounts of information, and due to the abundant data the microblog can offer, not only an algorithm with higher accuracy can be done but also can expand the scope of the number of users of large answer corresponding, rather than simply confined to the Q&A community’s internal answer user.This paper presents an automatic question classification method based on semantic imformation, which is to enhance the problem in using semantic template. First, step is to build a basic feature space according to the database, then each classification can be presented by the vector in the feature space. For a new question, first is to present the question by the original vector based on the word frequency, followed by semantic template extracting keywords and giving higher weights of keywords. Then computing the similarity of feature words of question and the feature words of the feature space by the WordNet, and the question is represented by he vector in the feature space. Finally through computing the similarity calculation between the vector of the question and vector of the classification, the classification of the question can be done. The experimental results show that the weighted and semantic mappings of keywords to improve the classification accuracy in a positive way.At the same time, this paper also studies algorithm of the microblog user answering ability. First the traditional mode of characterization of microblog users based on the PageRank algorithm is introduced, then sources of these methods and the existing problems is analyzed. Then it analyzes the defects of PageRank algorithm on this problem and deficiency, followed by analysis of the behavior of user on microblog which has constructed the interaction between chain graph with weight, put forward the preliminary improvement PageRank algorithm. The finally experimental results show that the PageRank algorithm with weight compared with common PageRank algorithm is a better way to characterize the users’answering capacity.