Research On Network Sensitive Text Classification Method Based On Knowledge Graph

With the advent of the wave of informatization,the security and stability of cyberspace are particularly important.However,there are all kinds of sensitive informations in cyberspace,which encourage violent terrorist acts,promote national splittism,and even promote the overthrow of state power,which has a huge negative impact on the control of public opinion and the construction of a harmonious society.Therefore,it is necessary to conduct sensitivity classification research on these sensitive informations with different sensitivity levels.Based on investigating the classification and filtering methods of sensitive texts at home and abroad,and combining the existing technology and algorithms,this paper conducts a research on the classification of sensitive texts based on Knowledge Graph.This topic first builds a knowledge Graph for network-sensitive information.The entire construction process includes four steps: crawling the data set,extracting entities and entity relationships,knowledge fusion and knowledge storage;then calculating the sensitivity value of the sensitive text based on the constructed Knowledge Graph.The calculation of the sensitivity value of the text is based on the frequency of the sensitive words contained in the text,the sensitivity of the word itself and other information.Based on the Knowledge Graph of sensitive information,a new influence factor δ is added,which is a sensitivity factor that is calculated based on the attributes of sensitive word entities in the graph,and then the RIPPLENET framework is introduced,in which the size of the ripple set centered on the sensitive entity is set to get the Sen value of the existing entity relationship in the set of sensitive words,thereby transforming the hierarchical level of sensitive text from the level of sensitive words to the level of semantic analysis;finally,a threshold value is set according to the sensitivity value obtained by comprehensive calculation to classify the sensitivity level of the text.Experimental results show that compared with the traditional method of sensitive text classification with sensitive words as the granularity,this method can effectively improve the accuracy and comprehensiveness of sensitive text classification,especially for sensitive texts with higher sensitivity levels,and the accuracy rate is improved obviously.