Among the various aspect-based sentiment analysis (ABSA) subtasks, aspect-level sentiment analysis (ASC) is useful in identifying the sentiment orientation of given aspect terms in a sentence. This study proposes the use of dependency types in a graph attention network to improve ASC performance. From the methodology viewpoint, the proposed method improves existing graph-based neural networks by considering dependency type information as an edge type to learn edge embeddings. This can better model the relationship among the aspect terms, opinion terms and other context words in the sentence and achieve more accurate graph propagation. As such, from the application viewpoint, the proposed method can be easily extended to other ABSA subtasks such as extract aspect terms, opinion terms and aspect triplets to further boost more fine-grained (aspect-level) sentiment applications.

Aspect-level sentiment classification (ASC) is designed to identify the sentiment orientation of given aspect terms in a sentence. Previous neural networks have used attention mechanisms to align context words with the appropriate aspect terms. Without considering syntactic dependencies, these models may erroneously focus on context words that are not related to the aspect terms. To address this issue, the graph convolution network (GCN) and the graph attention network (GAT) are proposed to build a graph based on the dependency parse tree, allowing the representations of context words to be propagated to the aspect terms according to their syntactic dependencies. However, these models consider all syntactic dependencies to be of the same type, and thus may result in inappropriate propagation of word representations in the graph. To further distinguish between the syntactic dependencies, this study proposes a syntactic graph attention network (SGAN) to incorporate the knowledge of depen...