Automatic Mining of Topics and Communities for Browsing in Web2.0

Web 1.0與Web 2.0有兩項主要差異：一)‧創作者個人資訊之揭露 (二)‧標籤的出現與否Web 2.0 環境中搜尋資訊的方法與 Web 1.0 的不同，有其優點，但也有些問題：一)‧只注重文件搜尋,而沒有排序人二)‧若文件與人沒有和query相同的tag, 這些文件與人將不會被找出。三) ‧由人們透過tag叢集起來的社群,是主題社群,而非社會性的社群。論文嘗試解決上述中傳統 Web 2.0 環境中對於搜尋的問題。們提出一個新的 Web 2.0 搜尋架構，並實做一個在 Web 2.0 環境中搜尋及瀏覽的系統，此系統能達到：一)‧文件擷取(二)‧人物擷取(三)‧主題之探勘(四)‧社群之探勘(五)‧同時標記主題與社群(六)使用者能在文件、人物、主題、社群、標籤之間做瀏覽。系統能輔助使用者快速透過超連結在 Web 2.0 環境中，針對上述五種物件進行探勘與搜尋，這種新型態的搜尋方式可以帶給使用者不同的觀點。過在 SIGIR、SIGGRAPH、SODA 蒐集到的資料集，在本系統架構上進行的實驗，證了本系統可在多種資料集上進行文件及人物擷取，並能有效的增進使用者社群網路中對主題及社群的理解。The two main differences between web 1.0 and web 2.0 environments are:1) author space and (2) tagging .Hence, the ways of searching in web 1.0 or web 2.0 environments are not necessarily the same. In web 2.0 environments there are some limitation which the searches may pose:1). Previous Web 2.0 search focuses on documents search,2). documents and people will not be found if they don’t have the same tags to the query,and 3). the communities clustered by the tags belong to topical community instead of social community.n this paper we have proposed a framework to handle such restriction, and a system is thus built for web 2.0 searching and browsing. Our system has the abilities to do the followings in the social networks:1). Retrieving documents and people.(2). Mining topics and communities.(3). Labeling the topics and communities.(4). Enabling the users to browse among documents, people, topics, communities and tags.ue to the mined information and hyperlinks through which the users can browse, the user''s comprehension among the structures of the social are extended; the proposed system presents a way of surfing the social networks where the users may have never experienced before.n the experiments we have evaluated the framework on the data sets crawled from SIGIR, SIGGRAPH, and SODA; we have found that our framework is effective in retrieving the documents and people from the social networks, and is able to improve the experiences and comprehension of the users when they browse through our system.摘要 ibstract iicknowledgement ivable of Contents vist of Figures viiist of Tables viiihapter 1 Introduction 1.1 Background 1.2 Motivations 3.3 Idea 5.4 Thesis Structure 7hapter 2 Related Work 8.1 Using Tags to Improve Search Performance 8.2 Community 9hapter 3 Proposed Approach 12.1 LSI Model 12.2 Documents Retrieval 17.3 People Retrieval 20.4 Topics Mining 23.5 Communities Mining 24.6 Labeling Topics& Communities 25.7 Discussion 27hapter 4 Experiment 29.1 Documents Retrieval 30.2 People Retrieval 39.3 Topics Mining 47.4 Communities Mining 51hapter 5 Conclusion 53.1 Summary of Contributions 53.2 Future Work 5