Emotional Analysis Of Short Text In Chinese Social Networks

The social platform short text is being produced at an explosive rate due to the rapid development of mobile Internet.However,traditional text analysis was difficult to deal with social platform short text since short txt is sparse,casual and variable with times.Besides,most of deep learning method was aimed at English corpus.How to analyze and extract the emotional tendencies of Chinese social network short text has immeasurable commercial and social value.This paper firstly analyzes the traditional text analysis model,which combines the dictionary method with machine learning model to train emotion classifier.The result shows that the dictionary method is not applicable to the short and casual text and traditional data preprocessing filters the emotional characters resulting in the loss of important emotional characteristics.This paper proposed a novel mechanism of Chinese social platform sentiment analysis based on convolutional neural networks with multi-dimensional features,which combined semantic features from word vectors with sentiment features from emoticons.This novel mechanism utilized convolutional neural networks to mine the deep correlation between features and labels.The problem that short text was hard to analyze is solved through mining multi-dimensional features and utilizing the abstract features extraction ability of convolutional neural networks.The experimental results showed that the sentiment analysis accuracy was relatively increased by 2.62% with the integration of emoticons.In addition,the accuracy and F measure are relatively improved by 21.29% and 19.20% respectively compared with machine learning model based on lexicon.