End-to-End Learnable Item Tokenization for Generative Recommendation

2024-09-09

arXiv.org

AI Summary

Existing generative recommender systems decouple item tokenization from recommendation modeling, leading to token–item misalignment and suboptimal performance. To address this, we propose ETEGRec, the first framework enabling end-to-end joint learning of item tokenization and generative recommendation. ETEGRec employs a dual encoder-decoder architecture and introduces two novel alignment losses: (i) sequence-item alignment loss, which enforces precise mapping between generated token sequences and ground-truth items; and (ii) preference-semantic alignment loss, which bridges user preferences with item semantics. Furthermore, it adopts an alternating optimization strategy to co-train a learnable tokenizer and a generative recommender. Extensive experiments on multiple benchmark datasets demonstrate that ETEGRec consistently outperforms both conventional sequential recommenders and state-of-the-art generative recommenders, achieving simultaneous improvements in recommendation accuracy and token-level interpretability of item identifiers.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• 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

Recently, generative recommendation has emerged as a promising new paradigm that directly generates item identifiers for recommendation. However, a key challenge lies in how to effectively construct item identifiers that are suitable for recommender systems. Existing methods typically decouple item tokenization from subsequent generative recommendation training, likely resulting in suboptimal performance. To address this limitation, we propose ETEGRec, a novel End-To-End Generative Recommender by seamlessly integrating item tokenization and generative recommendation. Our framework is developed based on the dual encoder-decoder architecture, which consists of an item tokenizer and a generative recommender. In order to achieve mutual enhancement between the two components, we propose a recommendation-oriented alignment approach by devising two specific optimization objectives: sequence-item alignment and preference-semantic alignment. These two alignment objectives can effectively couple the learning of item tokenizer and generative recommender, thereby fostering the mutual enhancement between the two components. Finally, we further devise an alternating optimization method, to facilitate stable and effective end-to-end learning of the entire framework. Extensive experiments demonstrate the effectiveness of our proposed framework compared to a series of traditional sequential recommendation models and generative recommendation baselines.

Problem

• Effective construction of item identifiers for generative recommender systems
• Unifying item tokenization and generative recommendation into one framework
• Enhancing alignment between item tokenizer and generative recommender

Innovation

• End-to-end learnable item tokenization framework
• Dual encoder-decoder architecture for unified recommendation
• Recommendation-oriented alignment strategy for optimization