Experimental Results for a Method to Discover Human Relationship Based on WWW Bookmarks
Masahiro Hamasaki, Hideaki Takeda · 2001
In this paper, we show how human network can be found and used on WWW. We proposed a system called kMedia that can assist users to form knowledge for community by showing shared topics networks (STN) among them. kMedia uses bookmarks as users’ knowledge because the structure of bookmarks and web pages registered in bookmarks reflect user’s interest. We conducted an experiment to know how kMedia can support users. One result is that folder recommendation is more effective than page recommendation. The other is that recommendation is more effective for people belonging to the same real communities than those to different communities. According to these result, we propose a new measurement called ”category resemblance” that is recommendation measurement based on resemblance of folder structures. This measurement shows higher than all other system generated parameters and human evaluation to detect human relationship.