Aspect-level sentiment analysis method and model based on BERT and aspect feature positioning model

Abstract

The invention relates to an aspect-level sentiment analysis method and model based on a BERT and an aspect feature positioning model, and the method comprises the steps: firstly, obtaining high-quality context information representation and aspect information representation through a BERT model, so as to maintain the integrity of text information; then, constructing an attention encoder based on a multi-head attention mechanism to learn interaction between body representation and context representation, integrating the relation between body words and contexts, and further distinguishing contributions of different sentences and aspect words to classification results; secondly, constructing an aspect feature positioning model to capture aspect information during sentence modeling, and integrating the complete information of the aspect into interactive semantics so as to reduce the influence of interference words irrelevant to aspect words and improve the integrity of aspect word information; and finally, fusing the context related to the target and the target important information, and predicting the probabilities of different emotion polarities by using an emotion prediction factor on the basis of the fused information. According to the method, the implicit relationship between contexts can be better simulated, information of aspect words is better utilized, interference of information irrelevant to the aspect words is reduced, and therefore higher accuracy and macro F1 are obtained.

Description

technical field [0001] The invention belongs to the technical field of aspect-level sentiment analysis, in particular to an aspect-level sentiment analysis method and model (ALM-BERT) based on BERT and an aspect feature location model. Background technique [0002] E-commerce is a booming industry with an increasing importance to the global economy. In particular, with the rapid development of social media and the continuous popularization of online social platforms, more and more users begin to express their own emotional comments on various online platforms. These reviews reflect the emotions of users and consumers, and provide a lot of valuable feedback information about the quality of goods or services for sellers and governments. For example: before buying a product, users can browse a large number of reviews about the product on the e-commerce platform to decide whether the product is worth buying. Similarly, governments and enterprises can directly collect a large n...

AI Technical Summary

Problems solved by technology

In addition, it is difficult for neural networks to capture the long-term dependencies between aspect words and contexts, resulting in the loss of valuable information

Images