融合卷积神经网络和注意力的评论文本情感分析

Commentary Text Sentiment Analysis Combining Convolution Neural Network and Attention

针对现有文本情感分析算法中网络模型输入单一,同时缺乏考虑相似文本实例对整体分类效果影响的现状,提出一种融合卷积神经网络和注意力的评论文本情感分析模型.首先,利用KNN算法得到加权文本矩阵,获得相似文本特征,使得分类特征更加丰富.然后,通过加权文本矩阵与原始文本矩阵构建注意力,捕获更多关键信息,使模型做出准确的判断.最后,使用双通道卷积神经网络模型对文本情感分类.本文在三个不同的数据集上进行大量实验,表明本算法可以有效利用文本特征间的依赖性,获取更多有用特征.同时根据准确率、召回率、精确率、F1值等衡量指标,表明本文所使用的模型相较于其他算法效果更优,实现了良好的分类性能.

In view of the current situation that the network model input is single in the existing text sentiment analysis algorithms and the influence of similar text instances on the overall classification effect is not considered,commentary text sentiment analysis combining Convolution Neural Network and Attention is proposed. Firstly,the weighted text matrix is obtained by KNN algorithm to obtain similar text features,which enriches the classification features. Then,the attention matrix is constructed by weighting the text matrix and the original text matrix to capture more key information so that the model can make accurate judgment. Finally,a double-channel Convolution Neural Network model is used to classify text sentiment. A large number of experiments are carried out on three different datasets,which show that the algorithm can effectively utilize the dependency between text features and obtain more useful features. At the same time,according to the accuracy rate,recall rate,accuracy rate,F1 value and other measurement indicators,it shows that the model used in this paper is better than other algorithms and achieves good classification performance.