Social Recommendation Technology Research Based On Open Source Search Engine

With the popularity of social networks, social media has become one of the main sources of information in the Internet. Microblog, also known as Weibo, as the most popular way to post personal information, has a significant influence in daily life. The weibo data posted by users can be seen as a parallel time line to user’s life, recording and reflecting their interest and preference at that time point. Therefore, how to effectively use the information from the social media, to extract the user’s interest model, and to recommendation information to users, exempting users from information overload, has become a more and more important research subject.This article firstly introduces the related concepts of recommendation system, analysis of the present research status at home and abroad, then introduces recommendation algorithm based on collaborative filtering and recommendation algorithm based on content model. With analysis of traditional similarity algorithms for recommendation algorithm based on collaborative filtering, a new asymmetric similarity correlation pearson(ASC-Pearson) similarity algorithm is proposed to cover the shortage of traditional similarity algorithms. In the experiment using the MovieLens data set, the new ASC-Pearson algorithm outperforms traditional similarity algorithms.As weibo limits the word count of post(140words), the analysis to weibo data is not the same with traditional long text analysis. In order to establish a better topic model with weibo data, this paper analyses the characteristics of weibo and divides weibo data into supervised sample and unsupervised sample by judging weibo contain topic tag or not. A new semi-supervised LDA topic model is proposed by optimizing the traditional LDA model with topic tag extraction method. In semi-supervised LDA, more prior information are used to get a better topic allocation result.With the semi-supervised LDA topic model extracted from weibo data, this paper proposes OWF-Max-Weibo algorithm to inference user interest model by analyzing its personal weibo information with order-weighted factor(OWF). In this way, the recommendation system is built based on user interest model to recommend topic-related articles, including news, blogs, etc. With three sina weibo user selected as test user, after manual analyzing and comparing their weibo content, the effect of the recommendation system is verified.