Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

LLMRec: Large Language Models with Graph Augmentation for Recommendation (2311.00423v6)

Published 1 Nov 2023 in cs.IR

Abstract: The problem of data sparsity has long been a challenge in recommendation systems, and previous studies have attempted to address this issue by incorporating side information. However, this approach often introduces side effects such as noise, availability issues, and low data quality, which in turn hinder the accurate modeling of user preferences and adversely impact recommendation performance. In light of the recent advancements in LLMs, which possess extensive knowledge bases and strong reasoning capabilities, we propose a novel framework called LLMRec that enhances recommender systems by employing three simple yet effective LLM-based graph augmentation strategies. Our approach leverages the rich content available within online platforms (e.g., Netflix, MovieLens) to augment the interaction graph in three ways: (i) reinforcing user-item interaction egde, (ii) enhancing the understanding of item node attributes, and (iii) conducting user node profiling, intuitively from the natural language perspective. By employing these strategies, we address the challenges posed by sparse implicit feedback and low-quality side information in recommenders. Besides, to ensure the quality of the augmentation, we develop a denoised data robustification mechanism that includes techniques of noisy implicit feedback pruning and MAE-based feature enhancement that help refine the augmented data and improve its reliability. Furthermore, we provide theoretical analysis to support the effectiveness of LLMRec and clarify the benefits of our method in facilitating model optimization. Experimental results on benchmark datasets demonstrate the superiority of our LLM-based augmentation approach over state-of-the-art techniques. To ensure reproducibility, we have made our code and augmented data publicly available at: https://github.com/HKUDS/LLMRec.git

PDF HTML Abstract

LLMRec: Enhancing Recommender Systems with LLMs and Graph Augmentation

Introduction

Recommender systems are an essential part of online services, helping users navigate through vast amounts of content by suggesting items of interest. Traditional methods have focused on analysing user-item interaction patterns, often extending their capabilities by incorporating side information to improve recommendation quality. However, these methods face significant challenges, including data sparsity and the quality of the side information used. To address these issues, we introduce a novel framework, LLMRec, which utilizes LLMs for graph augmentation in recommendation systems. This approach aims to tackle the limitations of sparse implicit feedback and low-quality auxiliary information by enhancing the interaction graph from a natural language perspective.

Methodology

The core of the LLMRec framework is to augment the recommendation process through three strategies:

  1. Reinforcing User-Item Interaction Edges: LLMs are employed to sample pair-wise training data, augmenting potential interactions based on the textual content, thus increasing effective supervision signals.
  2. Enhancing Item Node Attributes: We generate additional attributes for items, leveraging the deep knowledge embedded in LLMs to improve the descriptiveness and relevancy of item features.
  3. Conducting User Node Profiling: By analyzing textual content related to user interactions, LLMs can generate enriched user profiles that better reflect individual preferences.

To maintain the quality of the augmented data, a denoised data robustification mechanism is introduced. It comprises noisy implicit feedback pruning and MAE-based feature enhancement, targeting the refinement of both augmented interactions and node attributes. These measures ensure the reliability of the LLM-generated content, preserving the fidelity of user preferences and item characteristics.

Theoretical Analysis and Practical Implications

From a theoretical standpoint, employing LLMs as augmentors addresses critical issues in recommender systems by providing a richer representation of user-item interactions and side information. Practically, the LLMRec framework significantly improves recommendation accuracy as demonstrated through extensive experiments on benchmark datasets. The framework not only contributes to advancing the state-of-the-art in recommendation systems but also opens avenues for leveraging the power of LLMs in understanding and predicting user preferences more accurately.

Future Directions

While LLMRec marks a significant step forward, several avenues remain open for further exploration. Integrating causal inference with LLM-based augmentation could offer deeper insights into user behavior, providing a robust foundation for counterfactual reasoning in recommendations. Furthermore, extending the framework to accommodate dynamic user preferences and contextual variations presents an exciting challenge, promising to enhance the personalization and adaptiveness of recommender systems.

Conclusion

In summary, the proposed LLMRec framework showcases the potential of leveraging LLMs for data augmentation in recommendation systems. By addressing the twin challenges of sparse interactions and low-quality side information, LLMRec sets a new benchmark for recommendation accuracy, reaffirming the importance of incorporating semantic understanding and contextual knowledge in modeling user-item relationships. As we look to the future, the intersection of LLMs and recommendation systems promises to yield innovative solutions tailored to the evolving landscape of user preferences and online content.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (61)
  1. TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation. arXiv preprint arXiv:2305.00447 (2023).
  2. Revisiting negative sampling vs. non-sampling in implicit recommendation. TOIS 41, 1 (2023), 1–25.
  3. Efficient neural matrix factorization without sampling for recommendation. TOIS 38, 2 (2020), 1–28.
  4. Heterogeneous graph contrastive learning for recommendation. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining. 544–552.
  5. Zheng Chen. 2023. PALR: Personalization Aware LLMs for Recommendation. arXiv preprint arXiv:2305.07622 (2023).
  6. Uncovering ChatGPT’s Capabilities in Recommender Systems. arXiv preprint arXiv:2305.02182 (2023).
  7. Graph neural networks for social recommendation. In ACM International World Wide Web Conference. 417–426.
  8. Magnn: Metapath aggregated graph neural network for heterogeneous graph embedding. In ACM International World Wide Web Conference. 2331–2341.
  9. Masked autoencoders are scalable vision learners. In CVPR. 16000–16009.
  10. Ruining He and Julian McAuley. 2016. VBPR: visual bayesian personalized ranking from implicit feedback. In AAAI, Vol. 30.
  11. Lightgcn: Simplifying and powering graph convolution network for recommendation. In ACM SIGIR Conference on Research and Development in Information Retrieval. 639–648.
  12. The curious case of neural text degeneration. arXiv preprint arXiv:1904.09751 (2019).
  13. MixGCF: An Improved Training Method for Graph Neural Network-based Recommender Systems. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining.
  14. Do LLMs Understand User Preferences? Evaluating LLMs On User Rating Prediction. arXiv preprint arXiv:2305.06474 (2023).
  15. A survey of recommendation systems: recommendation models, techniques, and application fields. Electronics 11, 1 (2022), 141.
  16. Bootstrapping user and item representations for one-class collaborative filtering. In ACM SIGIR Conference on Research and Development in Information Retrieval. 317–326.
  17. Text Is All You Need: Learning Language Representations for Sequential Recommendation. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining.
  18. GPT4Rec: A Generative Framework for Personalized Recommendation and User Interests Interpretation. arXiv preprint arXiv:2304.03879 (2023).
  19. Knowledge Graph Contrastive Learning Based on Relation-Symmetrical Structure. IEEE Transactions on Knowledge and Data Engineering (2023), 1–12. https://doi.org/10.1109/TKDE.2023.3282989
  20. Learn from relational correlations and periodic events for temporal knowledge graph reasoning. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval Conference on Research and Development in Information Retrieval. 1559–1568.
  21. Reasoning over different types of knowledge graphs: Static, temporal and multi-modal. arXiv preprint arXiv:2212.05767 (2022).
  22. Structure Guided Multi-modal Pre-trained Transformer for Knowledge Graph Reasoning. arXiv preprint arXiv:2307.03591 (2023).
  23. Contrastive self-supervised sequential recommendation with robust augmentation. arXiv preprint arXiv:2108.06479 (2021).
  24. Augmenting sequential recommendation with pseudo-prior items via reversely pre-training transformer. In ACM SIGIR Conference on Research and Development in Information Retrieval. 1608–1612.
  25. Ilya Loshchilov et al. 2017. Decoupled weight decay regularization. In ICLR.
  26. Parallel Knowledge Enhancement based Framework for Multi-behavior Recommendation. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management. 1797–1806.
  27. Hierarchical Projection Enhanced Multi-Behavior Recommendation. In Proceedings of the 29th ACM SIGACM SIGKDD Conference on Knowledge Discovery and Data Mining Conference on Knowledge Discovery and Data Mining. 4649–4660.
  28. Coarse-to-fine knowledge-enhanced multi-interest learning framework for multi-behavior recommendation. ACM Transactions on Information Systems 42, 1 (2023), 1–27.
  29. Pytorch: An imperative style, high-performance deep learning library. Conference on Neural Information Processing Systems 32 (2019).
  30. Aleksandr Petrov and Craig Macdonald. 2022. Effective and Efficient Training for Sequential Recommendation using Recency Sampling. In Recsys. 81–91.
  31. Learning transferable visual models from natural language supervision. In ICML. PMLR, 8748–8763.
  32. Representation Learning with Large Language Models for Recommendation. arXiv preprint arXiv:2310.15950 (2023).
  33. SSLRec: A Self-Supervised Learning Library for Recommendation. arXiv preprint arXiv:2308.05697 (2023).
  34. BPR: Bayesian personalized ranking from implicit feedback. arXiv preprint arXiv:1205.2618 (2012).
  35. GraphGPT: Graph Instruction Tuning for Large Language Models. arXiv preprint arXiv:2310.13023 (2023).
  36. Heterogeneous graph masked autoencoders. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 37. 9997–10005.
  37. Knowledge Distillation on Graphs: A Survey. arXiv preprint arXiv:2302.00219 (2023).
  38. Graph neural prompting with large language models. arXiv preprint arXiv:2309.15427 (2023).
  39. Learning mlps on graphs: A unified view of effectiveness, robustness, and efficiency. In The Eleventh International Conference on Learning Representations.
  40. Denoising implicit feedback for recommendation. In WSDM. 373–381.
  41. Neural graph collaborative filtering. In ACM SIGIR Conference on Research and Development in Information Retrieval. 165–174.
  42. Rethinking the Evaluation for Conversational Recommendation in the Era of Large Language Models. arXiv preprint arXiv:2305.13112 (2023).
  43. Counterfactual data-augmented sequential recommendation. In ACM SIGIR Conference on Research and Development in Information Retrieval. 347–356.
  44. Contrastive meta learning with behavior multiplicity for recommendation. In Proceedings of the fifteenth ACM international conference on web search and data mining. 1120–1128.
  45. Multi-Modal Self-Supervised Learning for Recommendation. In ACM International World Wide Web Conference. 790–800.
  46. Multi-Relational Contrastive Learning for Recommendation. In Proceedings of the 17th ACM Conference on Recommender Systems. 338–349.
  47. Hierarchical user intent graph network for multimedia recommendation. Transactions on Multimedia (TMM) (2021).
  48. Contrastive learning for cold-start recommendation. In ACM MM. 5382–5390.
  49. Graph-refined convolutional network for multimedia recommendation with implicit feedback. In MM. 3541–3549.
  50. MMGCN: Multi-modal graph convolution network for personalized recommendation of micro-video. In MM. 1437–1445.
  51. Self-supervised graph learning for recommendation. In ACM SIGIR Conference on Research and Development in Information Retrieval. 726–735.
  52. Multi-modal Graph Contrastive Learning for Micro-video Recommendation. In ACM SIGIR Conference on Research and Development in Information Retrieval. 1807–1811.
  53. CAMUS: Attribute-Aware Counterfactual Augmentation for Minority Users in Recommendation. In ACM International World Wide Web Conference. 1396–1404.
  54. Are graph augmentations necessary? Simple graph contrastive learning for recommendation. In ACM SIGIR Conference on Research and Development in Information Retrieval. 1294–1303.
  55. Where to go next for recommender systems? id-vs. modality-based recommender models revisited. In ACM SIGIR Conference on Research and Development in Information Retrieval.
  56. LightFR: Lightweight federated recommendation with privacy-preserving matrix factorization. ACM Transactions on Information Systems 41, 4 (2023), 1–28.
  57. Recommendation as instruction following: A large language model empowered recommendation approach. arXiv preprint arXiv:2305.07001 (2023).
  58. Latent structure mining with contrastive modality fusion for multimedia recommendation. TKDE (2022).
  59. Mining Latent Structures for Multimedia Recommendation. In MM. 3872–3880.
  60. Causerec: Counterfactual user sequence synthesis for sequential recommendation. In ACM SIGIR Conference on Research and Development in Information Retrieval. 367–377.
  61. Multi-level cross-view contrastive learning for knowledge-aware recommender system. In ACM SIGIR Conference on Research and Development in Information Retrieval. 1358–1368.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (9)
  1. Wei Wei (424 papers)
  2. Xubin Ren (17 papers)
  3. Jiabin Tang (15 papers)
  4. Qinyong Wang (11 papers)
  5. Lixin Su (15 papers)
  6. Suqi Cheng (17 papers)
  7. Junfeng Wang (175 papers)
  8. Dawei Yin (165 papers)
  9. Chao Huang (244 papers)
Citations (109)
View on Semantic Scholar