基于混合神经网络模型的民生监督文本分类方法OA

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.