Judicial Case Determination Methods Based on Event Tuple

Judges have to consider the motives of the defendant’s side and the sequence of actions of all judicial subjects in the process of sentencing. Text analysis methods based on word vectors and deep neural networks, although they give statistically better classification results, cannot explain the causality of the actions of various subjects in the case logic. In this paper, we propose an event semantic mining algorithm that attempts to make judicial decisions from the causal logic. The method identifies the behavioral subjects in judicial documents through an entity extraction algorithm and extracts the subjects’ core behavior and motivation to achieve the construction of the underlying event tuple. By calculating the event tuple weights between different categories of cases, combined with a heap sorting algorithm, an event semantic tree is constructed for each case. Finally, a set of event tuple coding algorithm is designed to input the event semantic tree into the deep forest algorithm for inference. The experimental results show that the proposed event semantic tree construction method and event tuple coding method not only have a good case decision accuracy. It also has a good logical explanation.