Network Sensitive Information Discovery and Empirical Research Based on Punishing BiGRU Model by Fusion Weight
[Purpose/Significance]The discovery of network sensitive information is of great significance for purifying cyberspace and maintaining social stability.The current research on network sensitive information discovery ignores the long-distance contextual semantics,which leads to poor discovery performance.This paper proposes a net-work sensitive information discovery method based on punishing BiGRU model by fusion weight of sensitive terms.[Method/Process]Firstly,statistical weight,category weight and sentiment weight of sensitive terms are obtained,and the three are fused to obtain the fusion weight of sensitive terms.Secondly,the weighted loss function of sensitive terms is constructed by using the fusion weight for punishing misidentification of the text containing sensitive terms on BiGRU model.Finally,the discovery of network sensitive information is realized based on the punished BiGRU model.[Result/Conclusion]Empirical results on a real dataset from Sina Weibo indicate that compared to the existing methods,the proposed method has a certain improvement in Precision,Recall and F1 value.
long-distance contextual semanticssensitive terms weightsensitive information discoveryBiGRU
NETL
NSTL
万方数据
