Text emotion recognition method based on LDA and BERT fusion improved model

Abstract

The invention discloses a text emotion recognition method based on an LDA and BERT fusion improved model, and the method comprises the following steps: (1) obtaining a social network text, and carrying out the preprocessing; (2) fusing semantic features and theme features of the text, and outputting a word vector matrix; (3) inputting the features into a bidirectional Transform encoder, connecting a Softmax layer improved by gradient optimization, and outputting a classification model; (4) official corpora are put into the classification model, parameters are finely adjusted, and the model is improved. And performing emotion recognition on the social network text by using the obtained final classification model to obtain a more accurate recognition result.

Description

technical field [0001] The invention relates to a text emotion recognition method based on an LDA and BERT fusion improved model, and belongs to the technical field of text data recognition. Background technique [0002] With the advent of the era of big data and the vigorous development of 5G networks, the Internet has gradually advocated a user-centric open architecture, and the release of network information has increasingly changed from "timely" to "real-time". Internet users are changing from receivers to publishers of information. As a platform that can easily publish and obtain information, social networks are attracting more and more users to express their opinions on news and facts, and publish emotional texts about their personal lives. Therefore, how to accurately, timely and effectively obtain the emotional information of social network texts is of great value. There are three common text sentiment analysis methods. They are sentiment analysis method based on ...

AI Technical Summary

Problems solved by technology

Now the neural network model commonly used for emotion classification is LSTM, which alleviates the gradient explosion problem of general RNN to a certain extent, but there are still some problems, such as low parallel computing efficiency and slow running speed, etc.

With the advent of the Transformer model, the BERT model based on the former has excellent performance in many NLP tasks, but due to the lack of input of large-scale emotional corpus in the pre-training stage, it still has certain bottlenecks in performing sentiment analysis tasks.

Images