Deep Recognition of Network Sensitive Information by Integrating Implicit Emotion and Topic Enhanced Distribution
[Purpose/significance]Network sensitive information is an important object of network ecological governance.It can effec-tively promote the process of cyberspace purification by accurately identifying sensitive information on the network.[Method/process]In order to improve the recognition performance of network sensitive information,this study proposes a deep recognition method for network sensitive information that integrates implicit emotion and topic enhanced distribution.First,this study utilizes the proposed IME method to measure the implicit emotion of sensitive words.The implicit emotion feature of network information is obtained by fus-ing the implicit emotion of sensitive words with the original emotion of the information.Then,this study improves the BTM model by combining the special characteristics of network sensitive information,and extracts the sensitive topic enhanced distribution of net-work information based on the improved model.Finally,this study adopts the attention mechanism to fuse implicit emotion feature and sensitive topic feature with traditional sensitive word feature and semantic feature to realize the deep recognition of network sensitive information.[Result/conclusion]The experimental results show that both implicit emotion feature and sensitive topic feature of net-work information can effectively improve the sensitive information recognition performance.Compared with the existing methods,the network sensitive information recognition method based on multi feature fusion has better performance.[Innovation/limitation]This study deeply explores the implicit emotion feature and sensitive topic feature of sensitive information,and integrates multiple features to achieve effective recognition of network sensitive information.This study provides theoretical support for network ecological gover-nance,but further research is needed to enhance the distribution of sensitive topics in information.
NETL
NSTL
万方数据
