融合图卷积神经网络和BiGRU-TextCNN的罪名预测模型

Charge Prediction Model Integrating Graph Convolutional Neural Network and BiGRU-TextCNN

随着人工智能的发展和司法信息化体系的构建,运用人工智能解决司法领域的需求成为近年来自然语言处理领域的研究热点.为了提升深度学习模型在罪名预测任务上的准确率,本文设想利用将图卷积神经网络提取文本之间的结构特征与传统的深度学习模型相融合,捕获更多潜在信息.因此本文提出了一种融合图卷积神经网络和BiGRU-TextCNN的罪名预测模型,实验结果表明在Cail2018数据集上对罪名预测问题有较好的效果.

With the development of artificial intelligence and the construction of judicial information system,the need of using artificial intelligence to solve the issues in judicial field has become a research hotspot in the field of natural language processing in recent years.In order to improve the accuracy of deep learning models in charge prediction tasks,this paper envisages the integration of structural features in extracted texts by graph convolutional neural networks with traditional deep learning models to capture more potential information.Therefore,this paper proposes a charge prediction model that integrates graph convolutional neural network and BiGRU-TextCNN,and the experimental results show a good effect on the charge prediction problem in Cail2018 dataset.