基于多重卷积循环网络舆情分析方法的研究

Research on public opinion analysis method based on multiple convolution cycle network

针对规模化、精确化网络舆情分析的需求,文中对文本情感的分析方法进行了研究。通过结合深度学习中卷积神经网络(CNN)和循环神经网络(RNN)的优点,提出了多重卷积循环网络(CRNN)。该网络既保留了CNN深层次、拟合能力强的特性,又引入RNN中的长短记忆单元(LSTM),提升网络对于长文本序列的分析能力。基于该网络,其对网络舆情的分析方法流程进行了设计。仿真结果表明,所提出的方法在标准数据集NLPCC2013上,准确率、召回率和F1值相较于RNN、CNN网络分别可以提升约6%、2%和2%。

To meet the needs of large-scale and accurate network public opinion analysis,this paper studies the text sentiment analysis method By combining the advantages of Convolutional Neural Network(CNN)and Recurrent Neural Network(RNN)in deep learning,a multiple Convolutional Recurrent Neural Network(CRNN)is proposed.The network not only retains the characteristics of CNN’s deep level and strong fitting ability,but also introduces the Long Short-Term Memory unit(LSTM)in RNN,which improves the network’s analysis ability for long text sequence.Based on the network,this paper designs the analysis method flow of network public opinion.The simulation results show that the accuracy,recall rate and F1 value of the proposed method can be improved by 6%,2%and 2%respectively compared with RNN and CNN on the standard data set NLPCC213.