Isomer: Transfer enhanced dual-channel heterogeneous dependency attention network for aspect-based sentiment classification

• Published in: Knowledge-based systems 2022-11, Vol.256, p.109879, Article 109879
• Main Authors: Cao, Yukun, Tang, Yijia, Du, Haizhou, Xu, Feifei, Wei, Ziyue, Jin, Chengkun
• Format: Article
• Language: English
• Subjects: Aspect-based sentiment analysis; Graph convolutional network; Heterogeneous graph; Knowledge graph
• ISSN: 0950-7051; 1872-7409
• DOI: 10.1016/j.knosys.2022.109879

Aspect-based sentiment classification aims to predict the sentiment polarity of a specific aspect in a sentence. However, most existing methods attempt to construct dependency relations into a homogeneous dependency graph with sparsity and ambiguity, which only considers one type of node and one type of edge, thus cannot cover the comprehensive contextualized features of short texts or consider any additional node types or semantic relation information. To solve those issues, we present a sentiment analysis model, Isomer, which performs dual-channel attention on heterogeneous dependency graphs incorporating external knowledge to integrate additional information effectively. Specifically, a transfer-enhanced dual-channel heterogeneous dependency attention network is designed in Isomer for modeling short texts by heterogeneous dependency graphs. These heterogeneous dependency graphs not only consider different types of information but also incorporate external knowledge. Experiments studies show that Isomer outperforms the-state-of-arts on diverse datasets. Furthermore, the results suggest that Isomer captures the importance of various information features to focus on informative contextual words. •Propose a novel model for aspect-based sentiment classification.•A heterogeneous dependency graphs-based dual-channel attention network is proposed.•Propose a KGs-based feature enhancement method to reduce ambiguity and sparsity.•Propose a global correlation introduction method to capture aspect-related contexts.