Research On Text Sentiment Analysis Method Based On Att-BiGRU-CRF Model

With the advent of the 5G era and the continuous rise of various e-commerce platforms,it is common for users to comment on goods and services on the Internet,and online comment data has shown an exponential growth trend.These text data usually contain huge commercial value and social value.Text sentiment analysis task can be used to mine useful information from massive data.Deep learning model is suitable for big data analysis and prediction,so it has been widely used in the field of natural language processing.Recently,Convolutional Neural Network(CNN),Long-term and Short-term Memory Neural Network(LSTM),Gated Recurrent Neural Network(GRU)and other technologies have been successively applied to text sentiment analysis tasks,and achieved good results with their respective advantages.At present,the popular text sentiment analysis method based on deep learning still has some problems,such as insufficient feature extraction of the existing sentiment analysis model,low accuracy of emotion classification,and unable to correct Chinese spelling errors in text data.In view of the above problems,this paper has done the following research work:(1)Correction of Chinese spelling based on N-gram model.The misspelling of Chinese words will cause great changes in the sentiment tendency of sentences.High-quality text data is the basis for improving the accuracy of sentiment analysis of Chinese texts.By correcting the Chinese text of the comment data,we can effectively restore the initial emotional tendency of the sentence and improve the accuracy of text sentiment analysis.In the existing sentiment analysis tasks,it is not considered to correct spelling errors in Chinese texts.In this paper,the N-gram model is used to realize the automatic error correction of Chinese text,and the Laplace data smoothing strategy is introduced to solve the problem of data sparsity in the model.Finally,experiments verify the feasibility of the model to realize Chinese automatic error correction.(2)Sentiment analysis of Chinese text based on Att-Bi GRU-CRF model.Aiming at the problems of insufficient feature extraction of existing models and low accuracy of sentiment classification in current sentiment analysis tasks,this paper proposes an Att-Bi GRU-CRF text sentiment analysis model.After preprocessing the Chinese text data such as Chinese error correction and Chinese word segmentation,the Chinese-word-vectors pre-trained word vector model is used to convert the text data into a vector form that the model can recognize and input into the model.First,use the bidirectional gated recurrent neural network to learn the context of sentences,and fully extract the semantic information and feature structure of the text;Secondly,the attention mechanism is introduced to calculate the weight of different words in the emotional tendency of sentences,and the feature extraction of key words is strengthened;Finally,use the conditional random field(CRF)is as classifier,and the correlation between contexts is also considered in the output layer to avoid illegal output and improve the accuracy of classification.This article sets up multiple sets of comparative experiments on the same data set,and compares the model proposed in this article with the traditional sentiment analysis model.Experiments prove that the Att-Bi GRU-CRF model has achieved good accuracy and F1 value in Chinese text sentiment analysis tasks,which verifies the effectiveness and practicability of this model.