Research On Text Sentiment Analysis Technology Based On Deep Learning

With the rapid development of information technology and Internet industry,more and more people express their comments and opinions on the network platform in the Web 2.0 era.Faced with the explosive growth of these commentary data,the method of extracting and utilizing the sentiment information has become a research hotspot,and text sentiment analysis technology came into being.Text sentiment analysis is the process of analyzing subjective texts to extract their emotional tendencies.So the subjective and objective analysis of text is the premise work of sentiment analysis.As an important branch of the field of natural language processing(NLP),text sentiment analysis is of great theoretical significance.With the introduction of word embedding,NLP techniques based on deep learning have developed rapidly.Faced with massive text data,the advantages of deep neural network in learning and expressing have been demonstrated.This paper focuses on the research of text sentiment analysis technology based on deep learning.The main contents are as follows:Firstly,we construct a text sentiment analysis model based on multi-head self-attention mechanism.We study the attention mechanism in deep learning.The self-attention mechanism can focus on the dependencies between words within the text.Therefore,we introduce the multi-head self-attention mechanism into the task of text sentiment analysis.In order to enhance the learning ability of our model,we combine the non-linear sublayer,bidirectional gated recurrent unit with our model.The experimental results show that the accuracy of the proposed model on the sentiment analysis task has been improved.Secondly,we construct a model based on linear gated convolutional network for the subjective and objective analysis which is the subtask of sentiment analysis.We analyze different gating mechanisms and then introduce linear gating mechanism into the convolutional neural network.We use multiple convolution kernels of different sizes to extract text features.The experimental results show that our model outperforms other models in subjective and objective analysis task.The main innovations and contributions of the thesis are as follows.In the field of text sentiment analysis,multi-head self-attention mechanism is introduced.And the model is improved by combining the nonlinear sub-layers,which improves the accuracy of the model.We propose a subjective and obj ective analysis model based on gated convolution,which combines linear gated mechanism with convolutional neural network.According to the characteristics of text data,we construct several convolution kernels with different sizes,which effectively improves the performance of our model.