Design And Implementation Of Sentiment Classification System Based On Deep Learning For Comment Text

With the rapid development of e-commerce,more and more people publish their opinions on each commodity quality,customer service attitude and logistics speed in the network.These data are valuable.These data can give customers some references when choosing a product,and also enable businesses to improve their services so as to better acquire users,and even find new needs and new business opportunities from these data.However,these data are usually unstructured text data,which is not convenient for automatic classification,analysis and processing of computers.It is a very important task to make a certain model of the text data so that it is easier to extract its characteristics.At present,the traditional way of sentiment analysis is mainly to manually build emotional dictionaries and select different feature selection methods for different fields,which is rather tedious,time-consuming and laborious.On the basis of analyzing and summarizing the merits and demerits of traditional sentiment analysis methods,this paper uses deep learning method to extract emotional features from texts.In particular,it is to use the trained convolution neural network to extract the features of the review,and then send it to the trained support vector machine to complete the emotional classification of the text.The main reason for this design is the combination of deep learning to automatically extract features and support vector machines with good classification performance.In addition,a contrast test of the method and traditional machine learning method is done to verify the correctness of the theory.The experiment shows that the accuracy of the method is improved by 2 percentile degrees in the emotional analysis of the review text.Finally,based on the idea of emotional analysis,an emotional analysis system for the review text of American takeout is developed,which can be used to crawl the review text of the American takeout system in real time and analyze it.It completes the collection and preprocessing of the corpus,the training of the word vector,the generation of the emotional model and the function of the emotion classification,and it can clearly show the result of the text emotion analysis,and it is convenient to use.The system mainly includes the following functional modules:1,the collection and preprocessing module of the corpus: the collection of the corpus needs to write a crawler,search the web page with the breadth first search algorithm,and then parse the network to extract the content and save it locally.After the corpus is collected,participle and participle are used to segment and stop word processing.2,the training module of the emotional model: using the word vector tool word2 vec to train the Chinese language of Wikipedia to obtain the word vector model,and then train the convolution neural network with the preprocessed corpus and mark good corpus to generate the emotional model and save it.3.The text emotional attitude analysis module: the text emotional attitude analysis first completes the training of the support vector machine,then uses the generated emotional model to process the predicted text,extracts the features of the text,and finally uses the trained support vector machine to classify the features to obtain the emotional polarity of the text.