Research On Incident Analysis Methods For Judicial Data

The event analysis task of judicial data is mainly used to identify the criminal behavior of the criminal subject and object in the case,thereby analyzing the criminal behavior relation of the case,and extracting the main behavior process of the case.Since the results of judicial trials are affected by criminal behaviors,the research of this task can effectively support judicial intelligent auxiliary research such as law recommendation and crime prediction.In the analysis of judicial events,this article uses the predicate head word as the trigger word of the judicial event,and identifies the judicial event by identifying the predicate head word and related elements.Then construct all event information that occurred in the case,and extract the key event information of the case.The main research works of this paper are as follows:(1)Research on the method of judicial events extraction.Due to the characteristics of judicial data,the introduction of the predicate head word makes it have multiple expressions in the judicial data,which causes the traditional method to not effectively obtain the semantic information of the predicate head word,which leads to the problem of difficulty in detecting judicial events.Therefore,this paper proposes a judicial event extraction model based on character and word interaction.The model uses the interactive representation of character information and word features,uses Cross-Bi LSTM to learn the context-dependent information of character and word interaction information in sentences,and calculates the optimal tag path of event elements through CRF.The₁ value on the judicial data set reached 84.41%,which exceeded the comparison method by 4.8%.(2)Research on the method of judicial critical event detection.Since the predicate head word does not distinguish between categories,in addition to detection the main behavioral event information of the case,it will also detect some noise events that have nothing to do with the case.Therefore,this paper proposes a judicial critical event detection model based on behavioral sequence.This model models all event information in the case and constructs a behavior sequence,uses GGNN to learn the semantic correlation features between events in the behavior sequence,and uses the Attention mechanism to filter the noise events in the case.The model has achieved good performance in large-scale judicial data,with an accuracy rate of 45.73%.