In recent years,classification and rating of social emergency event have attracted more and more attentions in emergency management.However,most of the current studies adopt the rule-based methods to identify the evidences for event judgement,and have troubles in event judgement due to the complexity of social emergency event composition and the flexibility of language description.Inspired by the idea of event extraction,this paper has proposed the event judgement method via BiLSTM(Bidirectional Long-Short Term Memory)and CRF(Conditional Radom Fields)based on event classification and evidence extraction.The social emergency event classification is carried out firstly,and then the event evidences are extracted based on event type.In the end,the rating of social emergency event is judged with the attention mechanism and the combination of event type and evidence.Experimental results show that the proposed method is more robust than rule-based ones,and effective in the social emergency event judgement.
NETL
