Research On Fine-grained Sentiment Analysis Of Chinese Text Based On Deep Learning

Sentiment analysis,as a task in the NLP field,calculates the corresponding sentiment polarity through text content,and has been applied in many fields of social platforms,e-commerce platforms,and public opinion analysis.Traditional sentiment analysis only analyzes the text as a whole and gives a single sentimental tendency.However,when there are multiple different evaluation aspects in the review text and the corresponding sentimental inclination is not uniform,the effect achieved cannot satisfy people today.Individual needs.Fine-grained sentiment analysis is a targeted judgment of the fine-grained level of sentiment in the text,which more comprehensively reflects the potential information of the review text,and provides decision support for consumers,businesses and platforms.This article focuses on the Chinese comment text to conduct in-depth research on the fine-grained sentiment analysis method based on deep learning.The main contents of this article are as follows:(1)The task of fine-grained sentiment analysis relies heavily on the extraction of local information.How to extract keywords from the text will be the focus of fine-grained sentiment analysis.For the extraction of fine-grained terms in the text,this paper proposes a word selection The word selection attention mechanism(WSAM)builds a fine-grained sentiment analysis model based on the word selection attention mechanism,which can effectively extract words related to the fine-grained sentiment analysis task from the text.Through comparative experiments on the AI Challenger 2018 public data set,and by comparing the model F1 value,the method in this paper has a good effect.(2)Aiming at the sentiment analysis problem in many aspects of the text,this paper proposes a fine-grained sentiment analysis model based on the LSTM-GateCNN(LSTM with Gated Convolutional Networks)network.Compared with the current popular fine-grained sentiment analysis model,this model does not require additional The external aspect information can identify and locate the aspect information in the text autonomously.In terms of the distinction between target words and context,and attention to important context words,use LSTM(Long Short-Term Memory)to learn continuous sequences and Regional CNN captures the characteristics of local information,combines them and incorporates the attention mechanism into the model to give weight to important information,which can highlight the content related to a specific aspect of the text while extracting emotional information.Furthermore,it is possible to simultaneously model the text aspect information and sentiment information,which can perform sentiment analysis on multiple aspects appearing in the text at one time,and achieve an effective fine-grained sentiment analysis effect.Experiments show that the classification accuracy of the model is greatly improved compared to other methods.(3)Designed and implemented a fine-grained sentiment analysis system for Chinese text.