Research on Dual-Channel Graph Neural Network Integrating Entity-Attribute Knowledge in Aspect-Level Sentiment Analysis

Aspect-level sentiment analysis (ABSA) plays a pivotal role in fine-grained sentiment understanding by determining the sentiment polarity of specific aspects within text. Traditional ABSA methods based on graph neural networks (GNNs) typically leverage internal syntactic and semantic dependencies to model the relationships between aspect and sentiment words. However, these methods often overlook the contribution of external knowledge, which can enhance the association between aspect terms and sentiment expressions, particularly in cases of ambiguous wording or limited annotated data. To address these challenges, this paper proposes a dual-channel graph neural network framework (EAK-DCGNN) that effectively incorporates entity-attribute knowledge from an external knowledge graph. Specifically, the model constructs a knowledge graph by extracting "entity-attribute-sentiment" relationships from external sources, while simultaneously building a text graph based on dependency syntax of the input text. Features from both graphs are learned in parallel through the dual-channel GNN architecture and subsequently integrated using a dynamic attention-based fusion mechanism. The fused representation is then fed into a classifier to predict aspect-level sentiment. Comprehensive experiments on benchmark datasets demonstrate that EAK-DCGNN achieves superior performance in both accuracy and F1 score compared with baseline models. Ablation studies further validate the significant contribution of the entity-attribute knowledge module and the dual-channel fusion strategy, highlighting their effectiveness in mitigating the issues of data sparsity and semantic ambiguity. This study provides a novel and practical approach for enhancing the robustness and precision of ABSA systems in real-world scenarios.