关键词： 卷积神经网络, 特定目标, dropout机制, 文本情感

Abstract: In the process of emotion analysis of specific target text,text sequence classification is limited by the labeling method,resulting in low accuracy and recall of analysis results,so a model of emotion analysis of specific target text based on convolution neural network is constructed.Specific target text sequences are analysed based on emotional differences,and the text feature matrix is used as input data to construct a convolutional neural network language model in the input layer,concatenating it into a part of speech sequence matrix.Segmented pooling captures different key features of text sequences and classifies and processes the extracted feature vectors.A dropout mechanism is added to complete sentiment classification of specific target texts,determine the importance information of each word in the text,and achieve sentiment analysis of specific target texts.The experimental results show that the accuracy of the model is higher than 84%,and the maximum recall rate is 87%,which can effectively achieve the emotional analysis of specific target text.

Key words: convolution neural network, specific objectives, dropout mechanism, text emotion