Intrinsic and Extrinsic Factor Disentanglement for Recommendation in Various Context Scenarios

Abstract

In recommender systems, the patterns of user behaviors (e.g., purchase, click) may vary greatly in different contexts (e.g., time and location). This is because user behavior is jointly determined by two types of factors: intrinsic factors, which reflect consistent user preference, and extrinsic factors, which reflect external incentives that may vary in different contexts. Differentiating between intrinsic and extrinsic factors helps learn user behaviors better. However, existing studies have only considered differentiating them from a single, pre-defined context (e.g., time or location), ignoring the fact that a user's extrinsic factors may be influenced by the interplay of various contexts at the same time. In this paper, we propose the Intrinsic-Extrinsic Disentangled Recommendation (IEDR) model, a generic framework that differentiates intrinsic from extrinsic factors considering various contexts simultaneously, enabling more accurate differentiation of factors and hence the improvement of recommendation accuracy. IEDR contains a context-invariant contrastive learning component to capture intrinsic factors, and a disentanglement component to extract extrinsic factors under the interplay of various contexts. The two components work together to achieve effective factor learning. Extensive experiments on real-world datasets demonstrate IEDR's effectiveness in learning disentangled factors and significantly improving recommendation accuracy by up to 4% in NDCG.

32 pages, 13 figures, 11 tables. Published on Transactions of Information Systems

References in corpus (26)

• A Simple Framework for Contrastive Learning of Visual Representations
• Learning deep representations by mutual information estimation and maximization
• DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank Systems
• Improving Graph Collaborative Filtering with Neighborhood-enriched Contrastive Learning
• BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer
• Neural Factorization Machines for Sparse Predictive Analytics
• Disentangling Long and Short-Term Interests for Recommendation
• LightGCL: Simple Yet Effective Graph Contrastive Learning for Recommendation
• Learning Disentangled Representations for Recommendation
• Are Graph Augmentations Necessary? Simple Graph Contrastive Learning for Recommendation
• Graph Convolution Machine for Context-aware Recommender System
• CL4CTR: A Contrastive Learning Framework for CTR Prediction
• MISSRec: Pre-training and Transferring Multi-modal Interest-aware Sequence Representation for Recommendation
• EulerNet: Adaptive Feature Interaction Learning via Euler's Formula for CTR Prediction
• Neural Graph Matching based Collaborative Filtering
• XSimGCL: Towards Extremely Simple Graph Contrastive Learning for Recommendation
• Self-supervised Learning for Large-scale Item Recommendations
• Multi-domain Recommendation with Embedding Disentangling and Domain Alignment
• Contrastive Learning with Positive-Negative Frame Mask for Music Representation
• BARS: Towards Open Benchmarking for Recommender Systems
• Detecting Beneficial Feature Interactions for Recommender Systems
• Causal Disentangled Recommendation Against User Preference Shifts
• Empowering Collaborative Filtering with Principled Adversarial Contrastive Loss
• Multimodal Pretraining, Adaptation, and Generation for Recommendation: A Survey
• CSGCL: Community-Strength-Enhanced Graph Contrastive Learning
• PMG : Personalized Multimodal Generation with Large Language Models