Aspect sentiment analysis method based on two-channel graph convolutional network

Abstract

The invention relates to the technical field of natural language processing, in particular to an aspect sentiment analysis method based on a two-channel graph convolutional network, which comprises the following steps: acquiring a to-be-analyzed target text; inputting the target text into a pre-trained aspect sentiment analysis model; firstly, a target text is subjected to text coding, and text semantic representation is generated; then constructing a text sequence graph and an enhanced syntactic dependency graph based on the target text, and modeling based on structural information of the two graphs; fusing the structure information of the two graphs in a mutual enhancement mode through a multi-layer CoGCN to generate a text structure representation; and finally, respectively obtaining sentence representations from the text semantic representation and the text structure representation, and combining to generate a final specific aspect sentence representation. And generating prediction probability distribution on the sentiment label based on the specific aspect sentence representation, namely, an aspect sentiment analysis result of the target text. According to the aspect sentiment analysis method, the accuracy and comprehensiveness of aspect sentiment analysis can be improved.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to an aspect sentiment analysis method based on a dual-channel graph convolutional network. Background technique [0002] With the rapid development of social media and e-commerce platforms, more and more Internet users are willing to post their own comments on a certain thing or product on the Internet, and these opinions contain the user's emotional information. Therefore, the analysis of emotionally inclined comments and comments on major platforms can bring many benefits. For example, consumers can learn more about product information based on product reviews on shopping websites; companies can modify user reviews by monitoring social media. Marketing information, brand positioning, product development; stockholders choose whether to buy stocks based on evaluation. Therefore, sentiment analysis is a text classification technology with great practical applica...

AI Technical Summary

Problems solved by technology

However, existing GCN-based methods usually only model one kind of structural information (i.e., syntactic dependency structure), while largely ignoring other rich structural information between words, such as words within a time window The continuous structure of , or the co-occurrence structure between words and words on the entire corpus, leads to poor comprehensiveness and accuracy of aspect sentiment analysis

Images