Research On Emotional Tendency Classification Of Comment Text Based On Deep Learning

In recent years,with the rapid development of the Internet,e-commerce also presents an increasingly prosperous situation.Users will evaluate the products they need after they buy them on the platform.These evaluations include users’ subjective emotions about the products.Therefore,emotional tendency classification and judgment of these evaluations play a key role in both academic and commercial fields.In the past research in this field,the classification of emotional tendency of text is generally applied to the way of emotional dictionary or traditional machine learning,but these methods are not suitable for the massive text generated in the current network environment and the complex semantic information contained in the text,so more and more scholars use deep learning technology to do it Research on the classification of text emotional tendency.In this paper,CNN in deep learning technology and LSTM,a variant of recurrent neural network,are applied to the study of tendency division.For the current classification methods,text description usually lacks the analysis of text context,ignores the features of words themselves,and ignores the traditional linguistic rules.In view of the above problems,the multi-channel CNN emotion classification model and the LSTM emotion classification model integrating linguistic rules are proposed respectively,and good results are achieved.The specific work is as follows:(1)In this paper,we propose a classification model of multichannel evolutionary network with attention mechanism(att-mcnn).The current text representation model lacks the analysis of the text context environment,and does not consider the characteristics of the words in the text itself,thus losing the emotional information contained in the words.In this paper,we propose a multi-channel CNN text sentiment classification model with attention mechanism.It makes full use of the emotional word information in sentences,takes into account the emotional tendency information contained in the structure of words,and integrates the structural information of words into the text representation through multiple channels in the input layer,which improves the accuracy of the model.Experiments show that the accuracy of multi-channel text input is 2.75% higher than that of traditional CNN text classification model.(2)This paper proposes a text sentiment classification model based on lr-lstm(linguistically regulated LSTM).The traditional linguistic rules are integrated into the long-term and short-term memory network,and the features of sentences in time series and the linguistic rules contained in sentence structure are fully integrated,which improves the accuracy of the model.The experiment shows that the text tendency classification model which integrates linguistic rules has achieved better results,and its accuracy is improved compared with the previous LSTM text classification model 4.1%.(3)On the basis of the above research,this paper also completed the establishment and implementation of the emotional attitude partition system based on deep learning.For this kind of classification system,the requirements are analyzed,and the corresponding functions and modules are designed to realize the automatic processing of data collection,text preprocessing and text sentiment classification.