Aspect-Aware Multi-Criteria Recommendation via Attention-Enhanced Graph Neural Networks
IEEE Access · 2026 · 14 · 86660-86676
DOI: 10.1109/access.2026.3699421 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Fatemeh Rakee und 2 weiteren Autor:innen. Die Arbeit nennt 57 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The exponential growth of user-generated content in online platforms has intensified the challenge of information overload, making effective recommendation systems (RSs) essential for personalized content discovery. Traditional RSs, often reliant on single overall ratings, fail to capture the multifaceted nature of user preferences. Multi-criteria recommender systems (MCRSs) address this limitation by allowing users to rate items across several dimensions, yet they still struggle to fully leverage the rich, unstructured information embedded in user reviews. In this study, we propose AAG-MCR, a novel graph-based MCRS that integrates advanced natural language processing, attention mechanisms, and graph neural network techniques to enhance both criterion-specific and overall rating predictions. Our approach automatically extracts aspect terms from user reviews and constructs token-level graphs that encode syntactic and semantic relationships. An attention mechanism highlights the most relevant parts of each review for specific evaluation criteria, enabling the model to focus precisely on criterion-relevant content and improving both interpretability and predictive power. A graph neural network processes these enriched graphs to predict ratings for each criterion, while a deep neural network aggregates these predictions to infer overall ratings. Extensive experiments on real-world datasets demonstrate that AAG-MCR significantly outperforms state-of-the-art baselines in both accuracy and interpretability. By effectively modeling fine-grained user preferences and leveraging the depth of review text through attention, our system delivers more accurate, robust, and aspect-aware recommendations, advancing the capabilities of MCRSs.
Zitieren
APA
Rakee, F., Hamzeh, A., & Mozafari, N. (2026). Aspect-Aware Multi-Criteria Recommendation via Attention-Enhanced Graph Neural Networks. IEEE Access, 14, 86660-86676. https://doi.org/10.1109/access.2026.3699421
Vancouver
Rakee F, Hamzeh A, Mozafari N. Aspect-Aware Multi-Criteria Recommendation via Attention-Enhanced Graph Neural Networks. IEEE Access. 2026;14:86660-86676. doi:10.1109/access.2026.3699421
BibTeX
@article{Rakee2026,
author = {Rakee, Fatemeh and Hamzeh, Ali and Mozafari, Niloofar},
title = {Aspect-Aware Multi-Criteria Recommendation via Attention-Enhanced Graph Neural Networks},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {86660--86676},
doi = {10.1109/access.2026.3699421}
}
EUVIMED