Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Knowledge-Aware Multi-Intent Contrastive Learning for Multi-Behavior Recommendation

Shunpan Liang, Junjie Zhao, Chen Li, Yu Lei

TL;DR

This work tackles the challenge of sparsity and divergent intents in multi-behavior recommendations by introducing KAMCL, a knowledge-graph aware framework that explicitly models user intents across behaviors. The method comprises a relation aware KG aggregation module, an intent generation module, and an intent based user multi-behavior interaction module, all jointly trained with two contrastive losses in addition to a BPR objective. By partitioning the knowledge graph by relation, applying per relation GNNs, and generating intents from relation information, KAMCL aligns user intents with item attributes to improve recommendation quality. Experiments on three real-world datasets demonstrate that KAMCL consistently outperforms state-of-the-art baselines, with notable gains on sparse data, highlighting the practical impact of integrating knowledge graphs and intent aware contrastive learning for multi-behavior recommendation.

Abstract

Multi-behavioral recommendation optimizes user experiences by providing users with more accurate choices based on their diverse behaviors, such as view, add to cart, and purchase. Current studies on multi-behavioral recommendation mainly explore the connections and differences between multi-behaviors from an implicit perspective. Specifically, they directly model those relations using black-box neural networks. In fact, users' interactions with items under different behaviors are driven by distinct intents. For instance, when users view products, they tend to pay greater attention to information such as ratings and brands. However, when it comes to the purchasing phase, users become more price-conscious. To tackle this challenge and data sparsity problem in the multi-behavioral recommendation, we propose a novel model: Knowledge-Aware Multi-Intent Contrastive Learning (KAMCL) model. This model uses relationships in the knowledge graph to construct intents, aiming to mine the connections between users' multi-behaviors from the perspective of intents to achieve more accurate recommendations. KAMCL is equipped with two contrastive learning schemes to alleviate the data scarcity problem and further enhance user representations. Extensive experiments on three real datasets demonstrate the superiority of our model.

Knowledge-Aware Multi-Intent Contrastive Learning for Multi-Behavior Recommendation

TL;DR

This work tackles the challenge of sparsity and divergent intents in multi-behavior recommendations by introducing KAMCL, a knowledge-graph aware framework that explicitly models user intents across behaviors. The method comprises a relation aware KG aggregation module, an intent generation module, and an intent based user multi-behavior interaction module, all jointly trained with two contrastive losses in addition to a BPR objective. By partitioning the knowledge graph by relation, applying per relation GNNs, and generating intents from relation information, KAMCL aligns user intents with item attributes to improve recommendation quality. Experiments on three real-world datasets demonstrate that KAMCL consistently outperforms state-of-the-art baselines, with notable gains on sparse data, highlighting the practical impact of integrating knowledge graphs and intent aware contrastive learning for multi-behavior recommendation.

Abstract

Multi-behavioral recommendation optimizes user experiences by providing users with more accurate choices based on their diverse behaviors, such as view, add to cart, and purchase. Current studies on multi-behavioral recommendation mainly explore the connections and differences between multi-behaviors from an implicit perspective. Specifically, they directly model those relations using black-box neural networks. In fact, users' interactions with items under different behaviors are driven by distinct intents. For instance, when users view products, they tend to pay greater attention to information such as ratings and brands. However, when it comes to the purchasing phase, users become more price-conscious. To tackle this challenge and data sparsity problem in the multi-behavioral recommendation, we propose a novel model: Knowledge-Aware Multi-Intent Contrastive Learning (KAMCL) model. This model uses relationships in the knowledge graph to construct intents, aiming to mine the connections between users' multi-behaviors from the perspective of intents to achieve more accurate recommendations. KAMCL is equipped with two contrastive learning schemes to alleviate the data scarcity problem and further enhance user representations. Extensive experiments on three real datasets demonstrate the superiority of our model.
Paper Structure (30 sections, 15 equations, 4 figures, 4 tables)

This paper contains 30 sections, 15 equations, 4 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Multi-behavior Recommendation
  4. Knowledge Graph-based Recommendation
  5. Contrastive Learning
  6. PROBLEM DEFINITION
  7. METHODOLOGY
  8. Relation-aware Knowledge-graph Aggregation Module
  9. Relation-aware knowledge graph segmentation
  10. Knowledge sub-graph propagation and aggregation
  11. Relation-based contrastive optimisation
  12. Intent Generate Module
  13. Intent representation learning
  14. Intent-Based User Multi-behavior Interaction Module
  15. User Intent Generate for Different behaviors
  16. ...and 15 more sections

Figures (4)

  • Figure 1: Illustrations of the difference between the conventional approaches and our approach for multi-behavior recommendation. (a) depicts the commonly used manner, and (b) depicts our method which considers the intents of the user when making different behaviors.
  • Figure 2: The framework of KAMCL
  • Figure 3: Impact of dimension $d$
  • Figure 4: Impact of intent number $|P|$