基于多头指针的司法事件检测方法

针对如何解决中文司法事件检测中触发词与上下文关系不足以判定事件实例、案件触发词表述相似以及同一个案件中多个触发词识别和分类模糊的问题，本研究提出一种基于多头指针的司法事件检测方法。首先，该方法将上下文信息和罪名特征融合作为输入，使用双向长短期记忆(Bi-directional Long Short-Term Memory，BiLSTM)网络捕获数据依赖关系，深入提取特征；然后，使用多头指针网络对字符间的依赖关系进行建模，有效捕捉句子中的触发词；最后，利用指针标注技术抽取触发词，实现司法事件的有效检测。在公开司法数据集LEVEN上实验验证该方法的有效性，其中微平均和宏平均的F1指标达到了87.53%和78.05%，优于现有模型。该方法不仅显著提高了事件触发词的识别精度，而且也增强了对复杂司法文本中事件上下文关系的把握能力。

Judicial Event Detection Method Based on Multi-head Pointer

This article aims to solve the problem that the relationship between trigger words and context in Chinese judicial event detection is not enough to determine the case instance,and the case trigger words are similar in expression,and the identification and classification of multiple trigger words in the same case are fuzzy. A judicial event detection method based on multi-head pointer is proposed. Firstly,the method integrates context information and crime features as input,and utilizes a Bi-directional Long Short-Term Memory (BiLSTM) network to capture data dependencies and extract features in-depth. Then,the multi-head pointer network is used to model the dependency relationship between characters,and the trigger words in the sentence are effectively captured. Finally,trigger words are extracted by pointer annotation technology to realize effective detection of judicial events. Experiments on the public judicial dataset LEVEN validate the effectiveness of this method,in which the F1 index of micro-average and macro-average reaches 87.53% and 78.05%,which is better than the existing model. This method not only significantly improves the recognition accuracy of event trigger words,but also enhances the ability to grasp the context relationship of events in complex judicial texts.