Aspect-Based Sentiment Analysis Using Gate Vectors and Graph Convolution Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sentiment analysis systems fail to accurately predict sentiment in complex sentences with multiple aspects or irrelevant terms, as they do not effectively utilize aspect terms and dependency trees, leading to inaccurate results.
Innovation Solution
The proposed solution involves generating gate vectors based on aspect terms and applying them to graph convolution networks (GCNs) to reduce the impact of irrelevant terms, using a multi-task loss function that includes prediction, divergence, and consistency losses to ensure accurate and diverse sentiment predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sentiment analysis systems are used, then the system is simple to implement, but the accuracy of sentiment prediction for complex sentences deteriorates
Solution Approach 1:
The patent segments the sentiment analysis task by identifying and isolating aspect terms within sentences, then analyzing sentiment toward each aspect separately using dependency trees and graph convolution networks. This segmentation allows the system to handle complex sentences with multiple aspects accurately by processing them as distinct units rather than treating the entire sentence as a single unit.
Solution Approach 2:
The patent introduces gate vectors as intermediary elements that modulate the influence of different words and aspects in the sentence. These gate vectors are computed based on aspect terms and applied to adjust the representation vectors in the graph convolution network, serving as a mediator that enables the system to focus on relevant aspects while suppressing irrelevant information, thereby improving prediction accuracy without requiring complete redesign of the entire system architecture.
2Adaptability or versatility
If the system processes complex sentences with multiple aspects, then the coverage and versatility improve, but the reliability of sentiment prediction deteriorates
Solution Approach 1:
The patent applies local quality by analyzing different parts of the sentence (different aspects) with different levels of detail and importance. Through dependency tree parsing and graph convolution, the system identifies which words and aspects are locally important for sentiment prediction and applies appropriate processing weight to each, allowing accurate sentiment analysis of complex sentences with multiple aspects while maintaining high reliability for each individual aspect prediction.
Data Source
AI summary
Systems and methods for natural language processing are described. One or more embodiments of the present disclosure generate a word embedding for each word of an input phrase, wherein the input phrase indicates a sentiment toward an aspect term, compute a gate vector based on the aspect term, identify a dependency tree representing relations between words of the input phrase, generate a representation vector based on the dependency tree and the word embedding using a graph convolution network, wherein the gate vector is applied to a layer of the graph convolution network, and generate a probability distribution over a plurality of sentiments based on the representation vector.