Analysis of the Passion Criminal Tendencies Based on the Topic-Emotion Model

With the penetration of the Internet into daily life, more and more crime clues could be discovered on the Internet. Therefore, this paper uses a new method to predict whether the instant message on the internet has the tendency of emotional crime. It is because that the single theme-based analysis of criminal tendency, which ignores the impact of emotions, may confound the declarative sentences and normal expression with statements that contain criminal tendencies. The method used in this study introduces the analysis of emotional fluctuations while analysing themes. In order to solve the problems above, the study adopts seq2seq model to analyse emotional fluctuation. Themes and the values of emotional fluctuation are put into linear SVM as input vector to judge whether the instant message has an emotional crime tendency. The average predication accuracy of the model reached 86%, which basically accords with the assumption made before the experiment. This model can not only be used to analyse emotional crime, but also to screen the radical opinions on the internet.