Method for analyzing emotion polarity of long text news public opinions

Abstract

The invention discloses a method for analyzing emotion polarity of long text news public opinions, which comprises the following steps: S1, collecting text data as a training sample, S2, performing data cleaning on the data of the training sample collected in the step S1, and processing special characters in the cleaned data of the training sample to obtain a data set; S3, segmenting the data set,and segmenting the training sample into a training set and a test set according to a proportion; S4, establishing a deep learning network based on the representation model and loading pre-training parameters; S5, building a long text sentiment polarity analysis network model; S6, modifying a training sample data structure; and S7, model training: adopting a stratified sampling and K-fold cross validation method to ensure that the sample proportion in the sample data set of each fold is consistent with the original data proportion during stratified sampling, storing the model result of each fold in the model with the highest score of the validation set, integrating the K-fold model to test the test set, and taking the average probability as the test result of the model.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, and in particular relates to a method for analyzing the emotional polarity of long-text news public opinion based on a Transformer structure. Background technique [0002] Judging the emotional polarity of news public opinion can be abstracted as a text classification problem in the field of natural language processing, that is, to judge the emotional polarity expressed by the news through the title and text. At present, the methods used for sentiment analysis are mainly divided into the following three types: [0003] 1. Method based on sentiment lexicon. The process of the traditional model method based on the emotional dictionary is to first construct the emotional dictionary, and use the dictionary to judge the emotional orientation and emotional strength of the words in the pre-analysis text, so as to realize the overall emotional classification of the text. Limitations...

AI Technical Summary

Problems solved by technology

[0007] The technical problem to be solved by the present invention is to provide a method for analyzing the emotional polarity of long-text news public opinion, which encodes the long text through a transformer (Transformer), and then analyzes the long text through a bidirectional threshold recurrent unit network (Bi-GRU). The method of text news public opinion sentiment polarity can solve the problem of low accuracy of traditional methods

Images