Pointing the way: active collaborative filtering
CHI '95 Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
Shilling recommender systems for fun and profit
Proceedings of the 13th international conference on World Wide Web
Usage patterns of collaborative tagging systems
Journal of Information Science
The Future of Social Networks on the Internet: The Need for Semantics
IEEE Internet Computing
Why we twitter: understanding microblogging usage and communities
Proceedings of the 9th WebKDD and 1st SNA-KDD 2007 workshop on Web mining and social network analysis
Interlinking the Social Web with Semantics
IEEE Intelligent Systems
Social networks and interest similarity: the case of CiteULike
Proceedings of the 21st ACM conference on Hypertext and hypermedia
Improving recommendations using WatchingNetworks in a social tagging system
Proceedings of the 2011 iConference
| Hi-index | 0.00 |
In this paper, I describe a better way to compute the information similarity between two users who are unilaterally connected. Unilateral relations are unidirectional connections and gain attention with the success of social tagging and microblogging systems. The relations are convenient and less bounded since people can make the connection without mutual agreement once they perceive that other users' information is worth. Using a social bookmarking data set, Delicious, I found that the traditional item unit-based similarity measures are not enough to show the common interests between a pair of unilaterally connected users. The similarity measure on the higher level such as metadata (root address of each URL) and macro-level tags (tags regardless of the annotated information item) showed better results. The significantly better results in metadata and macro-tag level similarity were also shown in the indirect relations, as well. I interpreted this result to mean that semantic information such as metadata and tags represent users' cognitive understanding of corresponding information. Therefore, in social tagging systems, it is better to match users not based on item-level similarity but based on the similarity on a higher level which embeds more semantic meanings.