Knowledge Adaptation from Large Language Model to Recommendation for Practical Industrial Application

2024-05-07

arXiv.org

AI Summary

Existing recommender systems overly rely on item ID embeddings while neglecting rich textual semantics, resulting in poor generalization and robustness. To address this, we propose LEARN, the first framework to integrate a frozen large language model (LLM) as a lightweight, fixed-text encoder—thereby infusing open-world linguistic knowledge into collaborative filtering signals. LEARN employs a dual-tower architecture to jointly model ID-based and text-based representations, and is specifically optimized for low-latency inference in industrial settings. Evaluated on six Amazon Review benchmarks, LEARN achieves state-of-the-art performance. Moreover, on large-scale production data and online A/B tests, it delivers significant improvements: +4.2% in click-through rate (CTR) and +3.8% in conversion rate (CVR), while maintaining high computational efficiency and practical deployability.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Contemporary recommendation systems predominantly rely on ID embedding to capture latent associations among users and items. However, this approach overlooks the wealth of semantic information embedded within textual descriptions of items, leading to suboptimal performance and poor generalizations. Leveraging the capability of large language models to comprehend and reason about textual content presents a promising avenue for advancing recommendation systems. To achieve this, we propose an Llm-driven knowlEdge Adaptive RecommeNdation (LEARN) framework that synergizes open-world knowledge with collaborative knowledge. We address computational complexity concerns by utilizing pretrained LLMs as item encoders and freezing LLM parameters to avoid catastrophic forgetting and preserve open-world knowledge. To bridge the gap between the open-world and collaborative domains, we design a twin-tower structure supervised by the recommendation task and tailored for practical industrial application. Through experiments on the real large-scale industrial dataset and online A/B tests, we demonstrate the efficacy of our approach in industry application. We also achieve state-of-the-art performance on six Amazon Review datasets to verify the superiority of our method.

Problem

• Recommendation Systems
• Product Description Information
• Performance Improvement

Innovation

• LEARN System
• Pre-trained Language Models
• Dual Tower Structure Optimization