Research Of Chinese Weibo Fine-Grained Sentiment Analysis Based On Att-RNN And CNN

Sentiment analysis refers to mining and analysis of emotional information in subjective texts.Fine-grained sentiment analysis divides the sentiments in the text more precisely to get more sentiment information.As the mainstream representative of the Internet social platform,Sina-Weibo has a large user base and contains a large amount of sentiment information.The fine-grained sentiment analysis of Chinese Weibo is the extraction and classification of the sentiment information contained in Weibo texts.Accurately identifying the hidden sentiment information in Weibo has certain significance both in theoretical research and practical application.This thesis will study the Chinese Weibo sentiment analysis,improve the traditional sentiment dictionary based method and apply it to sentiment analysis.Base on deep learning algorithm,this thesis applies Attention-Recurrent Neural Network and Convolutional Neural Network into Chinese Weibo fine-grained sentiment analysis.The main research contents of this thesis are as follows:1 This thesis analyzes and improves the sentiment analysis method based on sentiment dictionary,updates the original sentiment dictionary,redefines the sentiment judgment rules.Compared to the method before improvement,the accuracy of the algorithm is greatly improved.2 Based on Attention mechanism and Recurrent Neural Network(RNN),this thesis applies the Att-RNN model to fine-grained sentiment analysis.The join of attention mechanism can learn the emotional features in the text better,remove the influence of irrelevant information,and improve the accuracy of fine-grained sentiment classification.This thesis compares and analyzes RNN and Att-RNN models with three different modules by applying it to vast experiments,and compares the model result by using different optimizers.The effects of pre-training words vectors are compared and analyzed,and different parameters of the model are selected.The experiments prove that the model shows good performance on Weibo fine-grained neural network.3 In this thesis,the Convolutional Neural Network(CNN)model is applied to Chinese Weibo fine-grained sentiment analysis.CNN has achieved great success in image processing and computer vision,but has less application in NLP.This thesis adapts the CNN model to sentiment analysis tasks,tests the feasibility of the application model,compare and analysis the influence of pre-training word vector,compare different parameters of the model.Meantime,joining new hand-labeled data to test the applicability and generalized ability of the model.Last,this thesis compares and analysis the accuracy and efficiency of Att-RNN model and CNN model.The research content of this thesis provides new methods to Chinese Weibo finegrained sentiment analysis.