基于混合神经网络模型的民生监督文本分类方法  

Text Classification Method of Livelihood Supervision Based on Hybrid Neural Network Model

作　　者：龙华 华才健 王琦标 徐尽悦 LONG Hua;HUA Caijian;WANG Qibiao;XU Jinyue(School of Computer Science and Engineering,Sichuan University of Science&Engineering,Yibin 644000,China)

机构地区：[1]四川轻化工大学计算机科学与工程学院,四川宜宾644000

出　　处：《四川轻化工大学学报（自然科学版）》2024年第2期49-56,共8页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

基　　金：四川省科技计划项目(2021JDRC0011);四川轻化工大学研究生创新基金项目(Y2022174)。

摘　　要：随着民生监督对信息化要求的逐渐提高,高效准确地识别民生监督文本可以帮助纪检监察部门及时搜集和跟踪事件并进行处理。针对民生监督文本分类困难的问题,提出了基于Mengzi模型融合BiLSTM、注意力机制和TextCNN的混合神经网络模型MBC,以提高民生监督文本分类的准确率。该模型首先使用预训练模型Mengzi得到富含丰富语义信息的词向量,后接并行的BiLSTM结合注意力机制网络和TextCNN网络,分别提取文本全局和局部特征,最后将全局与局部特征进行融合,实现对民生监督文本的准确分类。实验结果表明,MBC模型在准确率、召回率和F1值均达到了89%以上,优于传统的文本分类模型,为民生监督文本分类问题提供了新的研究思路。With the gradual increase of the requirements of livelihood supervision on informationization,efficient and accurate recognition of livelihood supervision text can help the disciplinary supervision department to collect and track the events and deal with them in time.Aiming at the problem of difficult text classification for livelihood supervision,a hybrid neural network model MBC based on the Mengzi model fusing BiLSTM,attention mechanism and TextCNN is proposed to improve the accuracy of text classification for livelihood supervision.The model first uses the pre-training model Mengzi to obtain word vectors rich in semantic information,followed by the parallel BiLSTM combined with the attention mechanism network and TextCNN network to extract global and local features of the text respectively,and finally the global and local features are fused to realize the accurate text classification for livelihood supervised.The experimental results show that the MBC model achieves more than 89%in accuracy,recall and F1 value,which is better than the traditional text classification model,and provides a new research direction for the problem of text classification for livelihood supervision.

关 键 词：民生监督文本分类 Mengzi模型 BiLSTM 注意力机制 TextCNN 混合模型MBC 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]