Research On Emotion Recognition Technology Of Microblog Based On Psychological Prewarning Model

In recent years, college students’mental illness has raised clearly. Damages caused by psychological problems happened more frequently than ever, such as suicide, which has seriously affected the growth of the students and the normal teaching management order of the university. More and more universities have begun to pay attention to students’mental health work, and focused more on preventing and screening for mental illness. Teachers engaged in college students’ mental health work, however, is very short. This contradiction need to be resolved. Social network, represented by Microblog, has become an important platform for gathering personal life trajectory and emotional information, which also opened a new direction for the current research. By collecting students’microblog information, we can identify the signs of risk, and help students to accept clinical screening and psychological counseling in the future. The current work is using of artificial recognition method, looking for a signal of crisis, which is such a heavy workload, and a large number of micro-blog content can easily be neglect and missed. Analyzing the content of one’s microblog by computer, especially those important students with high possibility of mental illness, can reduce the intensity of artificial work of recognition, so we can improve the efficiency of screening and the accuracy of prewarning.There are two difficulties in Micro-blog emotion analysis:on one hand, emotional change is a very complex process, which needs to build an emotional model to effectively simulate; on the other hand, such short text of the Micro-blog has unstandardized rules of grammar, so traditional machine learning methods and knowledge base has difficulties in extracting a characteristic value as well as a poor effect identification. Thus, a complete, professional emotional dictionary is needed in the process of identifying emotions in the text. The recognition process also involves segmentation, part of speech tagging, syntactic analysis and other fields of study. Due to what we’ve mentioned above, the article carry out the work of the following three aspects:(1) A multi-level psychological prewarning model based on personality, mood and emotion space is designed in the thesis. The model has fully taken into account the human cognitive factors of personality, temperament, external stimuli, so that can simulate the occurrence of human emotions and the process of change. Therefore, the current emotional state of a specific person can be inferred.(2) The techniques of getting and analyzing one’s Microblog text is researched in the thesis, including improving the traditional reptiles technology, designing microblog spider model, automatically collecting microblog from a specific character; analyzing characteristics of the recognition of microblog emotion, improving emotional dictionary-based text analysis algorithm, constructing the emotional element model and its inference rules. Thus, the emotion of the text analysis process is simplified as the reasoning and statistic of those emotional elements. Meanwhile, microblog psychological dictionary (MPED-dictionary) is built, as well as the emotion dictionary, to act as an important basis for extracting emotional elements.(3) We design and implement a microblog emotion prewarning system based on a model of psychological prewarning. The system will meet the needs of the current research because it can automatically collect the information on the microblog from a specific person on campus, and select those who has the most possibility of having a psychological crisis by analyzing his microblog emotion.