Article: Enhancing cross-border e-commerce English text classification using graph neural networks and transfer learning Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.39 pp.22 - 36 Abstract: Cross-border e-commerce platforms frequently feature English text characterised by mixed terminologies and informal syntactic structures, posing significant challenges for conventional classification models due to sparse labelled data. To address these limitations, this study introduces a novel classification framework that synergistically integrates GNNs and transfer learning. Specifically, a heterogeneous text graph incorporating word-document relationships is constructed to capture semantic dependencies, followed by the implementation of a domain-adaptive transfer mechanism to mitigate data sparsity through knowledge migration from related domains. Experimental evaluations on publicly available datasets, including Amazon Review and AliExpress, demonstrate that the proposed method achieves an accuracy of 92.7%, outperforming the BERT baseline by 4.5 percentage points. Furthermore, it significantly enhances classification efficacy in critical scenarios such as marketing content analysis and post-sale complaint resolution. This research advances cross-domain e-commerce text analytics by providing robust solutions for data-scarce environments. Inderscience Publishers - linking academia, business and industry through research

Title: Enhancing cross-border e-commerce English text classification using graph neural networks and transfer learning

Authors: Peng Du; Yanping Jiang

Addresses: Department of Digital Commerce, Yantai Engineering and Technology College, Yantai 264000, China ' Department of Digital Commerce, Yantai Engineering and Technology College, Yantai 264000, China

Abstract: Cross-border e-commerce platforms frequently feature English text characterised by mixed terminologies and informal syntactic structures, posing significant challenges for conventional classification models due to sparse labelled data. To address these limitations, this study introduces a novel classification framework that synergistically integrates GNNs and transfer learning. Specifically, a heterogeneous text graph incorporating word-document relationships is constructed to capture semantic dependencies, followed by the implementation of a domain-adaptive transfer mechanism to mitigate data sparsity through knowledge migration from related domains. Experimental evaluations on publicly available datasets, including Amazon Review and AliExpress, demonstrate that the proposed method achieves an accuracy of 92.7%, outperforming the BERT baseline by 4.5 percentage points. Furthermore, it significantly enhances classification efficacy in critical scenarios such as marketing content analysis and post-sale complaint resolution. This research advances cross-domain e-commerce text analytics by providing robust solutions for data-scarce environments.

Keywords: graph neural networks; GNNs; migration learning; cross-border e-commerce; text categorisation; domain adaptation.

DOI: 10.1504/IJICT.2025.149785

International Journal of Information and Communication Technology, 2025 Vol.26 No.39, pp.22 - 36

Received: 17 Jun 2025
Accepted: 08 Jul 2025
Published online: 12 Nov 2025 *

Title: Enhancing cross-border e-commerce English text classification using graph neural networks and transfer learning

Keep up-to-date