Article: Reasoning application of knowledge graph fusion-enhanced graph neural networks in English reading comprehension Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.65 pp.26 - 44 Abstract: This study addresses reasoning deficiencies in English reading comprehension by proposing a dual-channel framework that fuses graph neural networks (GNN) with knowledge graphs (KG). By integrating semantic relationships with GNN node feature transfer, the model significantly outperforms BERT-base and traditional GNNs. Experimental results on 10,000 texts show 89.7% accuracy in multiple-choice questions and 83.5% in logical relationship recognition. Ablation studies confirm the KG constraints reduce overfitting by 35% and improve efficiency by 20%. This research is the first to couple entity embedding with dynamic aggregation, providing interpretable reasoning paths. The model excels in complex long-text scenarios, offering a quantitative evaluation tool for English teaching through multi-dimensional metrics. Inderscience Publishers - linking academia, business and industry through research

Title: Reasoning application of knowledge graph fusion-enhanced graph neural networks in English reading comprehension

Authors: Dongjie Li; Xueqin Gong

Addresses: Department of Foreign Languages, Lyuliang University, Lishi 033001, China ' Department of Foreign Languages, Lyuliang University, Lishi 033001, China

Abstract: This study addresses reasoning deficiencies in English reading comprehension by proposing a dual-channel framework that fuses graph neural networks (GNN) with knowledge graphs (KG). By integrating semantic relationships with GNN node feature transfer, the model significantly outperforms BERT-base and traditional GNNs. Experimental results on 10,000 texts show 89.7% accuracy in multiple-choice questions and 83.5% in logical relationship recognition. Ablation studies confirm the KG constraints reduce overfitting by 35% and improve efficiency by 20%. This research is the first to couple entity embedding with dynamic aggregation, providing interpretable reasoning paths. The model excels in complex long-text scenarios, offering a quantitative evaluation tool for English teaching through multi-dimensional metrics.

Keywords: knowledge graph; graph neural network; GNN; English reading comprehension; reasoning enhancement; semantic fusion.

DOI: 10.1504/IJICT.2026.154187

International Journal of Information and Communication Technology, 2026 Vol.27 No.65, pp.26 - 44

Received: 24 Dec 2025
Accepted: 29 Mar 2026
Published online: 15 Jun 2026 *

Title: Reasoning application of knowledge graph fusion-enhanced graph neural networks in English reading comprehension

Keep up-to-date