The Research Of Identification Of Sensitive Text Based On Sentiment Analysis For Triple-play

With the deepening of the process of promoting the triple play and the continuesimprovement of the blend between different network, channels of informationdissemination become more and more diverse. Sex, violence, reactionary separatist areparts of bad information which don’t meet requirement of countries. Dissemination way ofthe bad information becomes more diverse, and their spread range becomes wider, withthe harm becoming more serious. In order to protect the health and effectiveness oftransfer content in the network, video, text and other image information in the networkmust comply with national laws, regulations and policies. With the development of tripleplay technology, content regulation should achieve the goal that the whole network can becontrolled to provide protection tube, and should achieve the implementation of contentplayback of reliable triple play through multi-level content, business audit, regulatory andsafety management.Sensitive text recognition system is based on HUSTRIM system framework, and it’san information systems for recognizing harmful texts. It contains three identificationprocess, which including the disabled word detection, text classification of sex andviolence, and improper political commentary detect, and improper political commentary isthe most important. The research process of improper political commentary isimplemented through sentiment analysis techniques. Analysis system of text Sentimenttendencies based on neural network is built by setting a reasonable neural network model,a detailed description of algorithms for text analysis for emotional tendencies in the neuralnetwork is presented. The algorithm re-training text Sentiment word polarity through theright words to get the reasonable size, so as to enhance the tendency to judge the accuracyof the text. Combined design goals and needs of HUSTRIM system, Apply analysis system fortext Sentiment tendency which is based on neural network to sensitive characterrecognition system. Results of experiments show that the system can already meet therequirement of initial recognition accuracy.