Visualization of Relationship between Social Bookmark Users
Visualization of Relationship between Social Bookmark Users Kyohei Hamada Advanced Course of Management Information Engineering, Ube National College of Technology, JAPAN
Introduction Web page recommendation system using social bookmarking Explosive spread of Web space makes our be difficult to obtain the required information. Web page recommendation systems Visualization of relationship between social bookmark users Based on collaborative filtering Conventional approach Our propose Today’s presentation Intersection between Web users Number of page in Web space <<
Folksonomy Assumption for tag The technique that classify Web pages based on the keyword, called “tag” ・ Tag expresses the property of individual Web page ・ Tag reflects the individual user’s thought for classification
Tagging by Dr. Park Innovation Graphics Literature Tagging by Dr. Hwang RFID Graphics Work Page 1 Page 2 Page 3 Tagging in Folksonomy Folksonomy allows to give multiple tags to a Web page.
Social Bookmarking Many user’s bookmarks are associated with each other through tags . When a Web page is bookmarked, arbitrary tags are allowed to be given. Web service that provides Internet users with a place to store, classify, annotate, and share Web resources. del.icio.us Margarine Hatena Example
Data model of social bookmarking Frequency which User1 has used a tag of “tag1” User1 User3 User2 Tag1 Tag6 Tag2 Tag3 Tag5 Page 1 Page 2 18 14 4 6 16 25 10 5 2 19
Procedure for Visualization (3)Visualization of relationship between users (1)Gathering social bookmark data (2)Estimation of similarity between users based on intersection of their’s own tags Visualization of the relationship between social bookmarking users, through tags given by users. Procedure Purpose
(1) Gathering Social Bookmark data Popular social bookmarking service in Japan. The gathered Social Bookmark data ・ 100 users ・ 54,540 tags (Distinct tags : 5,599 tags) Hatena Social Bookmarking
(2)Similarity measure Tanimoto coefficient , extension of Jaccard coefficient, is used. , Feature vectors (consisted of frequency of each tag) A , B ： ： Users
(3)Visualization Node : User Edge : Similarity between users 0 . 1 < sim ( A,B ) ≤ 0 . 2 0 . 2 < sim ( A,B ) ≤ 0 . 4 sim ( A,B ) > 0 . 4 Thin and blue solid line Solid and yellow line Thick and red solid line : : : (ex User1 User3 User2 User4 0.30 0.70 0.15 0.20
Conclusions Visualization of relationship between users in social bookmarking (through tags). Future issues ・ The elimination of commonly used tags ・ Adoption of similarity based on a set of entries (URLs) Emphasize each user’s interest Comparison between tag-based and entry based similarity evaluation. We could understand the relationship between users intuitively.