Relational Graph Convolutional Networks for Sentiment Analysis

Document Type : Research Paper

Authors

Department of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran.

Abstract

With the growth of textual data across online platforms, sentiment analysis is essential for deriving insights from user-generated content. While traditional approaches and deep learning models have shown promise, they often cannot capture complex relationships between entities. In this paper, we propose using Relational Graph Convolutional Networks (RGCNs) for sentiment analysis, which provide better interpretability by modeling dependencies between data points represented as intercon- nected nodes in a graph structure. We demonstrate our method’s effectiveness through pre-trained language models such as BERT and RoBERTa with RGCN architecture on product reviews from Amazon and Digikala datasets and analyze the resulting performance. Our experiments underscore the strength of RGCNs in capturing relational information for sentiment analysis tasks.

Keywords



Articles in Press, Accepted Manuscript
Available Online from 24 May 2025
  • Receive Date: 08 February 2025
  • Revise Date: 10 May 2025
  • Accept Date: 22 May 2025