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

With the rapid development of Internet technology,the Internet has become the main channel for people to communicate and obtain information.The information on the Internet is mostly displayed in the form of short texts.How to quickly and effectively extract the information in short texts is a very valuable topic.The traditional short text classification method not only relies too much on the construction of sentiment dictionary,but also requires a lot of labor cost for feature extraction.By analyzing these issues and using the characteristics of short texts,a short text sentiment classification model based on LSTMTextCNN is designed to improve the traditional classification methods.In the course of the subject research,through reading a large number of journal documents,the development history and research status of short text classification were introduced,and the related technologies in short text classification were elaborated and compared.In the text vector representation stage,the optimized TF-IDF method is used to weight the feature vector,which improves the shortcomings of the traditional Word2vec word vector representation and accelerates the convergence speed of the loss function in the classification model.Design a short text sentiment classification model based on LSTMTC.The feature vector obtained after training of the model retains the local and global features in the original text.The experimental comparison of this model with other models fully verifies the model's performance Effectiveness.When designing the system,the B/S architecture pattern was adopted to design and implement a short text emotion classification system.The functional modules implemented by the system include a data acquisition module,a data preprocessing module,an emotion classification module,and a page display module,among which the emotion classification module applies a short text emotion classification model based on LSTM-TC.The system finally displays the results of short text sentiment classification and the statistical results of short text segmentation.The short text sentiment classification model designed in this way has an accuracy rate of 93%through comparative experiments,which is more than 1%higher than other models such as SVM and TextCNN.It can be seen that this model has better performance in short text classification tasks.This paper verifies that the short text sentiment classification system has good performance and complete functions through the system test,and has reached the goal specified by the requirements.