A joint event extraction model for psychological evolution process analysis of telecommunication network fraud victims

Telecommunications network fraud, as the fastest growing criminal crime at present, has seriously affected social security and stability. It is an effective way to deal with such crimes from the perspective of victim risk prevention, and it is necessary to develop technical methods for victim risk analysis. This paper proposes a risk mining and psychological analysis model of victims based on event joint extraction model, starting from the practical logic of the fraud process. Firstly, elements extraction of the fraud process is achieved using RoBERTa-BiLSTM-CRF. Then, argument classification is carried out using a multi-head attention mechanism to accomplish the task of event extraction in the fraud process. Finally, based on the knowledge obtained from event extraction and frequent patterns, a deep analysis of the psychological evolution process of victims at each stage is conducted. The verification on empirical data reveals that the precision, accuracy, and F1 value of event extraction by the event joint extraction module are the highest, reaching up to 82.6%, 81.3%, and 81.9% respectively, enabling high-precision extraction of events from telecommunication fraud text data. Based on this, the fraud process can be discovered and achieved a more comprehensive, authentic, and dynamic analysis of victim psychological evolution mechanisms.