Syntactic and Semantic Dependency Graph Convolutional Network for Aspect-Based Sentiment Analysis

Aspect-based Sentiment Analysis (ABSA) is a fine-grained sentiment classification task aimed at identifying the sentiment polarity of specific aspects in a sentence. Currently, Graph Convolutional Networks are being widely used in sentiment analysis, effectively modeling the connections between aspects and opinion words and capturing structural information from syntactic dependency parsing. However, due to the informal expressions and complexity of sentences, existing methods face issues with word co-occurrence frequency. To address these problems, this paper proposes a Syntactic and Semantic Dependency Graph Convolutional Networks (SSDGCN) model, which combines syntactic and semantic dependency relations while considering the complementarity between syntactic structure and semantic relevance. The model employs a dual-channel approach and incorporates a self-attention mechanism to learn the syntactic dependency tree and semantic dependency tree separately, and then fuses the features from the two Graph Convolutional Networks for sentiment classification. Experimental results on two public datasets demonstrate that combining syntax and semantics using Graph Convolutional Networks for aspect-level sentiment analysis is beneficial.