Novel Social-graph Representation to Extract Hidden Common Relations between People

摘要

Predicting potential demands of people, which the people have not recognized or can not express precisely by themselves, can be useful for services which provide content pieces for users. Social graphs, which represent relationship between people, are used for predicting people demand in the current internet-based services. However, these graphs cannot represent such relationships that two users reside in common communities or common places. In this paper, therefore, we propose to represent not only a person but also other things including a community and a social event, together as a node in a social graph. Our representation allows us to estimate who share a potential interest with a given person. We validate our representation using a real relational dataset.