Detecting user preference on microblog

Abstract

Microblog attracts a tremendous large number of users, and consequently affects their daily life deeply. Detecting user preference for profile construction on microblog is significant and imperative, since it facilitates not only the enhancement of users' utilities but also the promotion of business values (e.g., online advertising, commercial recommendation). Users might be instinctively reluctant to exposure their preferences in their own published messages for the privacy protection issues. However, their preferences can never be concealed in those information they read (or subscribed), since users do need to get something useful in their readings, especially in the microblog application. Based on this observation, in this work, we successfully detect user preference, by proposing to filter out followees' noisy postings under a dedicated commercial taxonomy, followed by clustering associated topics among followees, and finally by selecting appropriate topics as their preferences. Our extensive empirical evaluation confirms the effectiveness of our proposed method. © Springer-Verlag 2013.