Research On Sentiment Analysis Of E-commerce Product Reviews Based On Deep Learning

The text of commodity reviews on the e-commerce platform is exploding,and it is of great significance to use the techniques of natural language processing and machine learning to automatically and efficiently analyze the text of the e-commerce platform.The current mainstream sentiment analysis methods mainly based on rule,machine learning and deep learning approach.With the development of big data technology and the diversification of language forms,deep learning has become a research hotspot in the field of natural language processing,and has made a major breakthrough in the field of sentiment analysis,therefore the sentiment analysis method based on deep learning is deeply studied.The main research contents are as follows:1.Aiming at the problem that the text representation feature of the existing crossdomain sentiment classification method ignores the sentiment information of important words and has negative transfer during the transfer process,a Convolution-BiLSTM based on attention mechanism(AC-BiLSTM)model is proposed to realize the knowledge transfer.Firstly,the text is represented by a low-dimensional dense word vectors;Secondly,after obtaining local context feature using convolution operation,the long dependence between the features is fully considered by bi-directional long short-term memory networks.Then,the degree of contribution of different words to the text is considered by introducing an attention mechanism,and in order to avoid the negative trsnsfer phenomenon in the transfer process,the regular term constraint is introduced in the objective function.Finally,the parameters from a model trained on the source domain product reviews are transferred to the target domain product reviews,and fine-tuned on the labeled data in a small number of target domains.The experimental results show that AC-BiLSTM model can effectively improve cross-domain sentiment classification performance.2.Aiming at the problem of ignoring word order information and context-dependent information in traditional deep memory networks,an aspect-level sentiment analysis method based on convolution-bidirectional minimum gate unit memory network(CNNBiMGU-MemNet)is proposed.Firstly,the context is represented by the Word2 Vec word vector model,and the high-dimensional original data is mapped into low-dimensional and continuous word vectors;Then,the word vector is input into the convolutional memory network and the bidirectional minimum gate unit memory network to obtain the order information of the words in the text and context-dependent information respectively.Finally,the output vector representation of the convolutional memory network and the bidirectional minimum gate unit memory network is combined and input it into the softmax layer for sentiment classification.The experimental results show that the CNNBiMGU-MemNet model is superior to other comparison models in accuracy and macroaverage F1 values,which can effectively improves classification performance.