Environmental Incidents Detection from Chinese Microblog Based on Sentiment Analysis.

Environmental incidents affect the stable development of economy and society. If environmental incidents can be detected timely through microblog, it will be possible to reduce risk factors and improve social stability. Microblog has now become an important platform for generating and propagating incidents on the Web, it is an ideal field to detect incidents. However, since the texts of microblog messages are very short and unstructured, it is a challenging task to detect incidents from microblogs. Due to the diversity of Chinese microblogs, the results identified by Dynamic Query Expansion DQE always contain a considerable number of unrelated events. This paper proposes to filter microblogs by calculating emotion values via Sentiment Analysis SA, thus to improve the accuracy of detection. Experimental results demonstrated that DQEi¾ź+i¾źSA can be more accurate and effective to detect environmental incidents.