Zusammenfassung
We present a ranking approach for Twitter documents that exploits social hashtagging behavior. We first map topics of user interest, represented by keywords, to a set of twitter hashtags that we use as query terms to retrieve twitter documents (tweets) based on tf-idf scores, with the additional restrictions that the documents retrieved should occur before the query timestamp. We show that this simple method performs significantly better than a disjunctive baseline based on the topic description. The performance achieved makes it specially attractive for information and collaborative filtering tasks, where a personalized lists of items (e.g., tweets) needs to be computed based on the user-item interactions in the system.
Nutzer