Blog Topic Mining And Application In Social Network Services

With the appearance of web2.0, a growing number of people arecoming to participate in to Social Network Services. They have become animportant medium for people to communicate ideas and share interests inrecent years. Research for user-generated content can not only helpenterprises understand the different trends of interest among public, butcan also improve various online services and user experience.Blogs published and shared by users in this virtual world are one of themain sources of user-generated information. Classifying these freestyleblogs can help understand user interests and assist applications such assearch and marketing. This article centers its research scope on blog. Twoaspects have been studied: topic classification of blogs and latent friendrecommendation based on blog content. The details are as follows:1) In this paper, we propose a new method of multi-label classificationfor Chinese blogs. By applying Dempster–Shafer theory on semantic wordsimilarity algorithms, we achieve automatic classification without use ofdifficult-to-obtain training sets. Experiments were conducted on real worlddata from RENREN.com, the biggest social network in China. Resultsshow that the proposed method achieves satisfactory performance inmulti-labeling real world SNS blogs as well as corpus.2) Since growing numbers of Internet users have been dissatisfiedwith the existing network of friends, they hope to find unfamiliar userswith common interests in social networks; however, existing friendrecommendation systems are mostly based on the existing ties between users. To address this problem, we propose a latent friend recommendationsystem based on user generated blogs. By analyzing user’s blog topics, weobtain the user’s interest distribution. We have also introduced time-sensitive factor to compensate user interest changes. A rough and refinedlatent friend recommendation is thus introduced. Experiments on Renrendata showed most of the recommendation results are quite satisfactory, andcomparison experiments showed favorable results of the refinementprocess.