LLM-Enhanced Hypergraph Learning for Review-Based Cross-Domain Recommendation

Sepideh Ahmadian, Mehrnaz Ayazi, Danial Ebrat, Luis Rueda, Dima Alhadidi
Proceedings of the The 39th Canadian Conference on Artificial Intelligence, PMLR 318:1130-1136, 2026.

Abstract

A major challenge in recommender systems is data sparsity. Cross-domain recommendation (CDR) addresses this issue by transferring knowledge from high-resource (HR) to low-resource domains (LR), but existing methods largely rely on user ratings that provide only implicit preference signals. In this work, we propose a review-based CDR framework that leverages Large Language Models (LLMs) to extract fine-grained product aspects and associated user sentiments from reviews, capturing explicit and nuanced user preferences. The extracted aspects are aggregated across source and target domains, and the relationships among users, items, and aspect-level features are jointly modeled using a hypergraph representation. In this formulation, each hyperedge explicitly connects a user, an item, and the corresponding extracted aspects, enabling a unified representation of their interdependent relationships. The resulting model is trained with a hypergraph neural network (HGNN) to enable effective preference transfer across domains. Experiments show that our approach significantly improves personalized recommendations in data-sparse settings, outperforming strong baselines while maintaining efficient knowledge transfer through shared semantic representations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v318-ahmadian26a, title = {LLM-Enhanced Hypergraph Learning for Review-Based Cross-Domain Recommendation}, author = {Ahmadian, Sepideh and Ayazi, Mehrnaz and Ebrat, Danial and Rueda, Luis and Alhadidi, Dima}, booktitle = {Proceedings of the The 39th Canadian Conference on Artificial Intelligence}, pages = {1130--1136}, year = {2026}, editor = {Bouzar-Benlabiod, Lydia and Leung, Carson}, volume = {318}, series = {Proceedings of Machine Learning Research}, month = {25--29 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v318/main/assets/ahmadian26a/ahmadian26a.pdf}, url = {https://proceedings.mlr.press/v318/ahmadian26a.html}, abstract = {A major challenge in recommender systems is data sparsity. Cross-domain recommendation (CDR) addresses this issue by transferring knowledge from high-resource (HR) to low-resource domains (LR), but existing methods largely rely on user ratings that provide only implicit preference signals. In this work, we propose a review-based CDR framework that leverages Large Language Models (LLMs) to extract fine-grained product aspects and associated user sentiments from reviews, capturing explicit and nuanced user preferences. The extracted aspects are aggregated across source and target domains, and the relationships among users, items, and aspect-level features are jointly modeled using a hypergraph representation. In this formulation, each hyperedge explicitly connects a user, an item, and the corresponding extracted aspects, enabling a unified representation of their interdependent relationships. The resulting model is trained with a hypergraph neural network (HGNN) to enable effective preference transfer across domains. Experiments show that our approach significantly improves personalized recommendations in data-sparse settings, outperforming strong baselines while maintaining efficient knowledge transfer through shared semantic representations.} }
Endnote
%0 Conference Paper %T LLM-Enhanced Hypergraph Learning for Review-Based Cross-Domain Recommendation %A Sepideh Ahmadian %A Mehrnaz Ayazi %A Danial Ebrat %A Luis Rueda %A Dima Alhadidi %B Proceedings of the The 39th Canadian Conference on Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Lydia Bouzar-Benlabiod %E Carson Leung %F pmlr-v318-ahmadian26a %I PMLR %P 1130--1136 %U https://proceedings.mlr.press/v318/ahmadian26a.html %V 318 %X A major challenge in recommender systems is data sparsity. Cross-domain recommendation (CDR) addresses this issue by transferring knowledge from high-resource (HR) to low-resource domains (LR), but existing methods largely rely on user ratings that provide only implicit preference signals. In this work, we propose a review-based CDR framework that leverages Large Language Models (LLMs) to extract fine-grained product aspects and associated user sentiments from reviews, capturing explicit and nuanced user preferences. The extracted aspects are aggregated across source and target domains, and the relationships among users, items, and aspect-level features are jointly modeled using a hypergraph representation. In this formulation, each hyperedge explicitly connects a user, an item, and the corresponding extracted aspects, enabling a unified representation of their interdependent relationships. The resulting model is trained with a hypergraph neural network (HGNN) to enable effective preference transfer across domains. Experiments show that our approach significantly improves personalized recommendations in data-sparse settings, outperforming strong baselines while maintaining efficient knowledge transfer through shared semantic representations.
APA
Ahmadian, S., Ayazi, M., Ebrat, D., Rueda, L. & Alhadidi, D.. (2026). LLM-Enhanced Hypergraph Learning for Review-Based Cross-Domain Recommendation. Proceedings of the The 39th Canadian Conference on Artificial Intelligence, in Proceedings of Machine Learning Research 318:1130-1136 Available from https://proceedings.mlr.press/v318/ahmadian26a.html.

Related Material