Dual-Enhancement Product Bundling: Bridging Interactive Graph and Large Language Model

2026-04-15

AI Summary

This work addresses the limitations of existing product bundle recommendation methods in cold-start scenarios and the difficulty large language models (LLMs) face in directly modeling user-item interaction graph structures. To bridge this gap, the authors propose a novel paradigm that integrates graph learning with LLM-based semantic understanding through a Dynamic Concept Binding Mechanism (DCBM), which efficiently translates graph structures into natural language prompts to align domain entities with LLM tokenization. By combining graph neural networks with LLMs via graph-to-text generation and semantic constraint modeling, the approach significantly enhances recommendation performance, outperforming state-of-the-art methods by 6.3%–26.5% on three benchmarks: POG, POG_dense, and Steam.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Product bundling boosts e-commerce revenue by recommending complementary item combinations. However, existing methods face two critical challenges: (1) collaborative filtering approaches struggle with cold-start items owing to dependency on historical interactions, and (2) LLMs lack inherent capability to model interactive graph directly. To bridge this gap, we propose a dual-enhancement method that integrates interactive graph learning and LLM-based semantic understanding for product bundling. Our method introduces a graph-to-text paradigm, which leverages a Dynamic Concept Binding Mechanism (DCBM) to translate graph structures into natural language prompts. The DCBM plays a critical role in aligning domain-specific entities with LLM tokenization, enabling effective comprehension of combinatorial constraints. Experiments on three benchmarks (POG, POG_dense, Steam) demonstrate 6.3%-26.5% improvements over state-of-the-art baselines.

Problem

• product bundling
• cold-start items
• interactive graph
• large language model
• collaborative filtering

Innovation

• Dual-Enhancement
• Graph-to-Text
• Dynamic Concept Binding Mechanism
• Product Bundling
• Large Language Model