Content Based Microblogger Recommendation

Microblogs are social networking systems, where users frequently contribute very short (micro) posts. Microbloggers follow others' posts via subscription. However, it is a challenge to determine microbloggers worthy of following, since their contributions are fragmented amidst numerous and various tiny posts. Furthermore, microposts typically include abbreviations, mispeled words, special tokens, and are not grammatically correct. Consequently, typical NLP techniques are often not very useful in determining what a user posts about. This work proposes a content-based, as opposed to friendship-based, microblogger recommendation model, where given an user query a ranked list of microbloggers is produced. This paper presents the recommendation model, a prototype implementation, an user evaluation.