Interest Similarity of Group and the Members: the Case study of Citeulike

In this Web 2.0 era, many social Web systems support group activities. Groups are centered on the utility of information usefulness. Users join the same group on the Web because they are interested in the same topic in terms of a community of interest or practice. Herein, we examine the information similarity in self-defined group networks and specifically address not only the similarities between the same group members, but also the similarities between a group and the members. Our study found that a pair of users who are the members of the same group share significantly higher similarity in their personal collection than other pairs who are not members of any of the same groups on all explored levels (items, metadata, and tags). Especially, the degrees of similarity on the metadata and tag levels are much larger than the item similarity. The degree of the similarities between a group and the members, however, is much higher than the similarities of the same group members. More than 40% of all users have collections which are at least 50% overlapped with their group’s collections. These results show that group is good source of information, but each member has his own specific information needs and it is rarely similar to other members. Another interesting property of information-sharing in groupbased networks is that the number of groups that a user joined has significantly positive correlation with the size of their personal collection. Lastly, some members play an active role in introducing interesting information to their groups and further, some other members were perfectly influenced by the group collection (100% matches with the group collection). Overall, our findings support that groups could be feasible for guiding users to useful information.