The Image Restoration Algorithm Based On Texture Synthesis

With the development of Internet, technology makes new media such as weibo become a new platform of information publishing and sharing. However, the transmission of information with a large amount of information, such as false, violence,reactionary and other information, is seriously affecting people’s normal life. Therefore,as soon as possible to establish a model of opinion mining for new media is particularly urgent.In this thesis, we study and apply the opinion mining technology of short text, such as weibo, focusing on weibo feature extraction and representation, topic discovery and tracking as well as sentiment analysis of topic. In terms of weibo feature extraction and representation, this thesis puts forward a kind of graph structure model based on semantic. The graph structure model maps the key words to the nodes of graph model,and maps the relations of interdependence syntactic to the edges of graph model. The model fully considers the semantic information of the text, so compared with traditional methods, the model effectively overcomes the defects of lack of semantic. In terms of topic discovery and tracking technology, the thesis uses the Single-Pass algorithm to find topics of weibo, and optimizes the algorithm to solve some problems. Then the thesis defines and calculates the influence of a single weibo and weibo topic, and completes the recognition and extraction of hot weibo topics. In terms of sentiment analysis of a weibo topic, this thesis firstly chooses the emotional words which have an obvious tendency of emotion, and then treats them as seed words, and expands the emotional word library through the Ci Lin. Then this thesis tries to eliminate the influence of context to emotional words through the interdependence syntactic analysis.Compared with the traditional opinion mining model of short text, the model which I put forward more truely shows the semantic information of texts. Then this thesis chooses the medium-size test corpus to do experiments and conducts relevant evaluation.The results show that the model has achieved a better effect.