Emotion classification method and device

Abstract

The invention relates to an emotion classification method and equipment. The method comprises the following steps: inputting an embedded representation corresponding to a target sentence into a feature information extraction network to obtain a sentence feature representation; inputting the sentence feature representation and the weight graph into a weight graph syntax information extraction network to obtain weight graph syntax information; inputting the sentence feature representation, the syntactic adjacency matrix and the semantic adjacency matrix into a semantic syntactic information integration network to obtain semantic supplementary syntactic information; according to the position of the target word in the target sentence, the weight graph syntax information, the semantic supplement syntax information and an emotion information extraction algorithm, obtaining a first emotion feature representation and a second emotion feature representation; and fusing the first sentiment feature representation and the second sentiment feature representation, and inputting the fused sentiment feature representation into a sentiment classification function to obtain a sentiment classification result of the target word. Compared with the prior art, the sentiment classification method and the sentiment classification device have the advantages that unnecessary noise can be reduced, semantics are supplemented into syntactic information, and sentiment classification accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to an emotion classification method and equipment. Background technique [0002] Sentiment classification is an important task in Natural Language Processing (NLP), and its purpose is to classify subjective texts with emotional color. Among them, attribute-level sentiment classification is a fine-grained sentiment classification method, which is different from traditional sentiment classification methods, and its purpose is mainly to analyze the sentiment polarity of target words in sentences. [0003] However, the current attribute-level sentiment classification methods mostly use the syntactic dependency tree corresponding to the sentence to obtain the syntactic dependence information in it, so as to use the syntactic dependence information and deep neural network to judge the emotional polarity of the target word. However, since the syntactic dependency tree ...

AI Technical Summary

Problems solved by technology

However, since the syntactic dependency tree contains connections between words that are not related to sentiment classification, unnecessary noise is introduced, and if the syntactic structure of the sentence is complex, it will also lose a lot during the transmission of the deep neural network. Key information that affects the accuracy of the final classification results

Images