An Attribute-Level Sentiment Analysis Method Based on Hierarchical Attention Mechanism and Gate Mechanism

Abstract

The invention discloses an attribute-level sentiment analysis method based on a hierarchical attention mechanism and a gate mechanism, aiming to make full use of the relationship between the context and the attribute words, so that the context and the attribute words are fully associated; the context and the attribute words are highlighted In order to improve the accuracy of attribute-level sentiment analysis; to enrich the information in the context and attribute word representation, the solution is to preprocess the comment corpus, and obtain the context word embedding matrix and attribute word embedding through the GloVe word vector index Matrix; input the context and attribute words into the GRU to obtain the hidden state of the context and the hidden state of the attribute words; obtain the context vector representation 1, the attribute word vector representation and the context vector representation 2 through self-attention; splicing to obtain the overall vector representation, and through classification The device obtains the distribution of the emotional polarity corresponding to the attribute word in the context, analyzes the difference, and adjusts the parameters in the model; it belongs to the field of emotional analysis in natural semantic processing.

Description

An attribute-level sentiment analysis method based on hierarchical attention mechanism and gate mechanism technical field The present invention relates to a kind of sentiment analysis method, specifically, is a kind of based on hierarchical attention mechanism and gate mechanism Attribute-level sentiment analysis method; belongs to the field of sentiment analysis in natural semantic processing. Background technique [0002] In recent years, with the development of the Internet, a large number of consumer platforms and social network platforms have gradually entered people's Life. Consumers often leave a comment about the pros and cons of the product after consuming on the consumer platform, and netizens will also leave a comment. view on an event. For businesses, these comments with emotional information, whether for enterprises or the government Words are of great value, and companies can understand the shortcomings of products by analyzing consumer reviews, so as t...

AI Technical Summary

Problems solved by technology

However, simple splicing cannot achieve the effect of identifying important words for attribute words in the context

Second, in most attribute-level sentiment analysis work, the results of the attention mechanism between context and attribute words are directly used as the expression of context and attribute words. This paper argues that this approach ignores the role of the original context and attribute word representation to a certain extent

Images