Tag-Enriched Multi-Attention with Large Language Models for Cross-Domain Sequential Recommendation

2025-10-10

AI Summary

This paper addresses two key challenges in cross-domain sequential recommendation (CDSR): modeling dynamic cross-domain user interest evolution and capturing coarse-grained item semantics. To this end, we propose an LLM-driven label-enhanced multi-attention model. Methodologically: (1) we leverage large language models to generate domain-aware, fine-grained semantic labels for items; (2) we fuse textual and visual modality features to construct enriched item representations; and (3) we design a label-guided multi-attention mechanism to jointly model users’ preference evolution paths across multiple domains. Extensive experiments on four large-scale e-commerce datasets demonstrate that our approach significantly outperforms state-of-the-art baselines. Results validate the effectiveness of synergistic semantic label enhancement and multimodal attention for cross-domain sequential modeling. The proposed framework establishes a novel paradigm for LLM-augmented, interpretable, and fine-grained cross-domain recommendation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Application Domains: Humanities & Computational Social Science
• Computer Vision: Multi-modal Vision

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Cross-Domain Sequential Recommendation (CDSR) plays a crucial role in modern consumer electronics and e-commerce platforms, where users interact with diverse services such as books, movies, and online retail products. These systems must accurately capture both domain-specific and cross-domain behavioral patterns to provide personalized and seamless consumer experiences. To address this challenge, we propose extbf{TEMA-LLM} ( extit{Tag-Enriched Multi-Attention with Large Language Models}), a practical and effective framework that integrates extit{Large Language Models (LLMs)} for semantic tag generation and enrichment. Specifically, TEMA-LLM employs LLMs to assign domain-aware prompts and generate descriptive tags from item titles and descriptions. The resulting tag embeddings are fused with item identifiers as well as textual and visual features to construct enhanced item representations. A extit{Tag-Enriched Multi-Attention} mechanism is then introduced to jointly model user preferences within and across domains, enabling the system to capture complex and evolving consumer interests. Extensive experiments on four large-scale e-commerce datasets demonstrate that TEMA-LLM consistently outperforms state-of-the-art baselines, underscoring the benefits of LLM-based semantic tagging and multi-attention integration for consumer-facing recommendation systems. The proposed approach highlights the potential of LLMs to advance intelligent, user-centric services in the field of consumer electronics.

Problem

• Captures domain-specific and cross-domain user behavioral patterns
• Generates semantic tags using Large Language Models for item enrichment
• Models complex user preferences across multiple e-commerce domains

Innovation

• LLMs generate semantic tags from item content
• Tag embeddings fuse with multi-modal item features
• Multi-attention models user preferences across domains