Topic Detection and Tracking (TDT) is one of the issue in the field of Natural Language Processing (NLP) from the beginning. It becomes a hot spot and many classic models and methods have been made currently, as it faces the information retrieval and extraction detecting on unknown topics and track exist ones. The paper reviews the development status of TDT research. The definitions of Topic, Story, Event and Activity are stated. TDT corpus and primary tasks in current research are introduced. The integral framework and key technologies including topic models, correlation calculation and clustering and classification, are proposed. The status of development of topic detection and tracking is concluded in the end.
展开▼
