Research On The Key Technology Of Text Sentiment Analysis Based On Deep Learning

The rapid increase in the scale of Internet users has led to the massive quantification of data scale,and higher requirements have been placed on the efficiency and capabilities of existing data processing technologies.At the same time,natural language processing and text processing technologies are facing unprecedented challenges.How to understand efficiently and accurately the rich emotions contained in massive text data is one of the key technical difficulties that need to be solved urgently.It can effectively serve upper-level applications,such as public opinion analysis,popular event prediction,etc.,so it is a key research issue that is generally concerned by industry and academia.one.The main research object of this thesis is text sentence analysis technology based on deep learning.First,we will introduce in detail text processing technology,traditional machine-based text computing methods,key technologies of deep learning networks,and related applications in natural language.This thesis proposes two algorithms.One is the sentiment analysis model based on LSTM,which includes an input module,encoder module,LSTM module,and output module.The evaluation on the real text sentiment data set Sentiment140 dataset shows that the proposed sentiment analysis model based on the LSTM is better than the existing classic models Text Bi RNN,Text Att Bi RNN,HAN,and CNN,and performs on different evaluation indicators Accuracy,Precision,Recall,and F1 better.The other algorithm is the sentence analysis model based on the attention mechanism.The model adds an intermediate interpretation layer.The interpretation layer of the algorithm has good text information interpretation capabilities,which improves the effect and interpretability of information in text sentiment calculation.The text interpretation layer is to capture the span information of the text,and finally assign a specific weight,thus avoiding the problem of excessive dependence on other external auxiliary models,thereby improving the classification effect of the model.Through experimental analysis,it is concluded that the interpretation layer of the sentiment analysis model based on the attention mechanism has a good performance in extracting the text.The encoder in the sentiment analysis model based on LSTM modules to obtain more context and information,and the sentiment analysis model based on the attention mechanism improves the model effect by adding an interpretation layer to the model.The sentiment analysis model based on the attention mechanism does not rely on additional model explanations and shows great model performance.