Abstract

Twitter can be viewed as a data source for Natural Language Processing (NLP)tasks. The continuously updating data streams on Twitter make it challenging totrace real-time topic evolution. In this paper, we propose a framework formodeling fuzzy transitions of topic clusters. We extend our previous work oncrisp cluster transitions by incorporating fuzzy logic in order to enrich theunderlying structures identified by the framework. We apply the methodology toboth computer generated clusters of nouns from tweets and human tweetannotations. The obtained fuzzy transitions are compared with the crisptransitions, on both computer generated clusters and human labeled topic sets.

 

Quick Read (beta)