JERARTOP: A New Topic Detection System

In this paper we present an on-line detection system, named JERARTOP, which goes beyond traditional detection systems, because it generates the implicit knowledge of a stream of documents. This knowledge is expressed as a taxonomy of topics/events, which is automatically built by the system in an incremental way. Moreover, the proposed detection system also annotates each detected topic using a set of predefined subjects, as well as it provides a summary for the topic. The experimental results demonstrate its usefulness and its effectiveness as a detection system.