부산대학교 도서관

In recent years, there has been a growing interest in utilizing dependency parsing with graph convolutional networks for aspect-based sentiment analysis. Dependency relations between words are used to construct graphs that integrate syntactic information into deep learning frameworks. However, most existing methods fail to consider the impact of different relation types between content words, which makes it difficult to distinguish important related words. Moreover, the semantic relationship between words can enhance the text understanding ability, which has been largely neglected in previous works. To address these limitations, in this paper, we propose a novel model named as SS-GCN. Our model automatically learns syntactic weighted matrix and leverages semantic information to obtain the text semantic representation, and an attention module is introduced to obtain the specific aspect-context hidden vectors. The model enhances the text representation ability from syntactic and semantic graph convolutional networks. We conducted comprehensive experiments on publicly available datasets to demonstrate its validity and effectiveness. The experimental results demonstrate that our model outperforms strong baseline models.