Topic detection and tracking oriented to BBS
Xiulan Hao, Yunfa Hu · 2010
Because topic detection and tracking (TDT) shares similar challenges with information retrieval, information filtering and information extraction in bursts of news stories, it has become a hot spot in the community of nature language processing. The TDT system oriented to BBS can detect and track the special event netizens paying close attention to and plays an important role in capturing public opinion. First, state of the art in topic detection and tracking is reviewed. Then a real-world application is given. In the system, a baseline model is given according to the characteristics of BBS. To alleviate “topic drifting” in TDT, an improved model based on the baseline model is proposed. The late reweighting of named entity (NE) is applied to the improved model to reallocate weight of NE features. Finally, experimental results on real data set are given.