Document-level sentiment classification method based on attention combination neural network

Abstract

The invention discloses a document-level sentiment classification method based on an attention combination neural network. The method is divided into two stages to obtain the characteristics of the document-level comments for sentiment classification: in the first stage, a two-layer bidirectional gating recurrent neural network is used for obtaining document matrix representation with compositionsemantics, and a double-layer attention mechanism is used for distinguishing the importance of different words and sentences; in the second stage, a two-dimensional convolutional neural network is used to sample more significant feature dependencies in the matrix representation obtained by the first stage and generate a high-dimensional representation for emotion classification, and in the secondstage, a convolutional attention module is used to focus on important features and suppress unnecessary features. According to the method, the problems that long text modeling is difficult to performand dependence among features is difficult to capture in an existing sentiment classification method are solved, and important information can appear in any position and composition part of a documentin a long text.

Description

technical field [0001] The invention belongs to the field of natural language processing, and in particular relates to a document-level sentiment classification method based on an attention combination neural network. Background technique [0002] Text sentiment classification is one of the most used natural language processing techniques in many fields, such as e-commerce websites, political trend analysis, online social networks, etc. Text sentiment classification is an important task in sentiment analysis. Traditional text sentiment classification methods include dictionary-based methods and corpus-based methods. Dictionary-based methods utilize existing word or phrase sentiment lexicons and some language rules to achieve sentiment prediction for documents, while corpus-based methods mostly rely on texts with annotated sentiment polarity to build classifiers. With the great success of deep learning, researchers began to use deep learning methods, such as convolutional n...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a document-level sentiment classification method based on the attention combination neural network, which is used to solve the problems that it is difficult to carry out long text modeling and capture the dependence between features in the existing sentiment classification method

Images