This paper presents an initial exploration into the feasibility of automatic topic detection in instant messaging applications. We have developed a prototype system that employs a set of preprocessing techniques, weighted scheme for word combination and suitable word classification and extraction of non-relevant words. The topic detection is based on the descriptive metadata for the selected keywords provided by an existing public knowledge base. In addition, the selected keywords have been linked with popular tweets to provide supplemental information. An exploratory user study has been conducted to gather some insights into the performance and usability metrics related to the proposed approach.
Real-time topic detection preprocessing techniques IM chat message characteristics