Research On Text Sentiment And Its Time Trend Based On Deep Learning

Text sentiment analysis is a branch of text classification,which applies in social media,e-commerce platforms and other fields.It has the value of studying people's emotional tendencies and attitudes towards entities such as products and services.Traditional sentiment analysis methods have problems such as the need to manually extract features and the inability to associate before and after semantic understanding.In this thesis,Bi-LSTM model is selected for text sentiment analysis.This model uses the RNN network structure to automatically obtain feature expression capabilities,and solves the problem that semantics cannot be correlated before and after.Aiming at the problem that Bi-LSTM cannot highlight the key information in the text,this thesis introduces Convolutional Neural Network and Attention Mechanism to improve its feature learning ability,and specifically makes the following three improvements: first,Convolutional Neural Network is introduced to extract local features of the text,and improve the feature expression ability of the model on the basis of the Bi-LSTM model;second,Attention Mechanism is introduced to assign different weights according to the importance of text information,so as to improve the problem that Bi-LSTM cannot highlight the key points of text information;third,Convolutional Neural Network and Attention Mechanism are introduced to extract the local features of the input information and assign different weights to the output information of the Bi-LSTM layer,so as to improve the problem that Bi-LSTM cannot highlight the key points of the text information and improve the model feature expression ability.The experimental results show that the improved Bi-LSTM model has improved accuracy,Marco-F1,and recall,which are increased by 1.57%,2.59%,and 2.84%,respectively,indicating that the improvement of the model is effective.The text has time attributes.Studying its emotional temporal trend can understand the changes of customers' overall emotions over time,and then understand the changes of customers' needs,and provide guidance for the development and improvement of new products of the enterprise.In this thesis,the ARIMA model is used to study the temporal trend of text sentiment,the model parameters are determined through the ADF test of sequence stationarity and the BIC criterion,and then the DW test is used to evaluate the model.Experiments show that the test results are within the range of residual non-autocorrelation,indicating that the model is effective and can be applied to the study of the temporal trend of text sentiment.