DALLRec: an effective data augmentation framework with fine-tuning large language model for recommendation

Abstract Recommender systems play an important role in modern digital platforms, but data sparsity has been one of the main challenges in this domain. Traditional solutions usually mitigate this problem by introducing side information, but often face the challenge of poor data quality. This can sign...

Full description

Saved in:
Bibliographic Details
Main Authors: Hongzan Mao, Baisong Liu, Xueyuan Zhang, Wei Liu, Zijing Wang, Zining Feng
Format: Article
Language:English
Published: Elsevier 2025-08-01
Series:Journal of King Saud University: Computer and Information Sciences
Subjects:
Online Access:https://doi.org/10.1007/s44443-025-00158-4
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract Recommender systems play an important role in modern digital platforms, but data sparsity has been one of the main challenges in this domain. Traditional solutions usually mitigate this problem by introducing side information, but often face the challenge of poor data quality. This can significantly hamper the accuracy of user preference modeling, thus affecting the recommendation effectiveness. Given the rapid development of large language models (LLMs) technology in recent years, with their rich semantic knowledge and powerful generative capabilities, we propose a new framework called DALLRec, which aims to achieve data augmentation by fine-tuning LLMs. The DALLRec framework effectively mitigates the data sparsity problem and significantly improves the overall performance of the recommender system through three core strategies: enhancing user-item interactions, enriching item attributes, and generating high-quality item summaries. We integrated DALLRec with multiple SOTA recommendation models and experimentally verified its effectiveness. The experimental results show that DALLRec can adapt to multiple recommendation models and significantly improve their recommendation performance, providing a new direction for further improving the accuracy of recommendation systems. Our code and data are available at https://github.com/zzerrrro/DALLRec .
ISSN:1319-1578
2213-1248