Research On Text Sentiment Analysis Algorithms Based On Deep Learning

In the era of big data,people leave a lot of information in the areas of business platform,movie community platform,and discussion platform.Text as the most common carrier of these data,if we can use natural language processing(NLP)technology to dig out the potential emotional attitudes of text,it will promote the development of network public opinion supervision,after-sales,automatic decision-making,and has very important practical application value.Traditional text sentiment analysis methods are mainly based on emotional dictionary and machine learning(ML).In the method based on emotional dictionary,whether it is constructed manually or automatically,there are problems of insufficient coverage and single application field.The Naive Bayesian(NB),Maximum Entropy(MAE)and Support Vector Machines(SVM)are common methods based on ML.In these methods,insufficient samples and difficult extraction of artificial features will seriously affect the final text sentiment analysis results.Since it was proposed,Deep Learning(DL)has attracted widespread attention in academic circles.Based on the excellent achievements of DL in the field of NLP,this paper proposes a text sentiment analysis algorithm based on DL.The main content of this paper is divided into the following four parts.Firstly,the Chinese text is used to cut sentences,word segmentation,stop words(punctuation marks,English characters),and the words are statistically built to establish a dictionary;then,the dimensional disaster and data sparseness problem for traditional word vectors are represented.This paper uses deep learning techniques such as embedding layer,word2 vec and glove to obtain low-dimensional,dense word vectors.The main contribution of this paper is to propose two algorithm models.For the Long-Short Term Memory(LSTM)neural network to learn only the historicalinformation of the text and ignore the future information,the LSTM is replaced by Bidirectional Long-Short Term Memory(BiLSTM)neural network,and combines the Convolutional Neural Network(CNN)to obtain the text features.A text sentiment analysis model based on CNN and BiLSTM is proposed.considering that different words in the text have different effects on the final text classification.This paper introduces a Multi-Head Attention mechanism(MHA).A text sentiment analysis algorithm model based on BiLSTM and MHA mechanism is proposed.Finally,the text has made multiple sets of comparative experiments on the proposed model.The former model uses Tan Songbo dataset and IMDB dataset.The latter model uses a commodity review data set on the e-commerce platform.The precision rate(p),recall rate(r),F1 scores(f1),and accuracy(acc)are used as criteria for evaluating the quality of the model.The experimental results show that the two deep learning models proposed in this paper further improve the performance of text sentiment analysis.