IRIT at TREC Microblog 2015

Abstract

This paper presents the participation of the IRIT laboratory (University of Toulouse) to the Microblog Track of TREC 2015. This track consists in a real-time filtering task aiming at monitoring a stream of social media posts in accordance to a user's interest profile. In this context, our team proposes three approaches: (a) a novel selective summarization approach based on a decision of selecting/ignoring tweets without the use of external knowledge and relying on novelty and redundancy factors, (b) a processing workflow enabling to index tweets in real-time and enhanced by a notification and digests method guided by diversity and user personalization, and (c) a step by step stream selection method focusing on rapidity, and taking into account tweet similarity as well as several features including content, entities and user-related aspects. For all these approaches, we discuss the obtained results during the experimental evaluation.