Review-Based Domain Disentanglement without Duplicate Users or Contexts for Cross-Domain Recommendation

TL;DR

This paper introduces SER, a review-based model for cross-domain recommendation that effectively disentangles domain-specific information without relying on shared users or contexts, improving robustness and scalability.

Contribution

The paper proposes a novel review-based disentanglement approach with a new optimization strategy and multi-domain encoding, advancing cross-domain recommendation methods.

Findings

Outperforms state-of-the-art methods in accuracy and robustness.

Effective in scenarios with no shared users or contexts.

Scalable and efficient for large-scale e-commerce data.

Abstract

A cross-domain recommendation has shown promising results in solving data-sparsity and cold-start problems. Despite such progress, existing methods focus on domain-shareable information (overlapped users or same contexts) for a knowledge transfer, and they fail to generalize well without such requirements. To deal with these problems, we suggest utilizing review texts that are general to most e-commerce systems. Our model (named SER) uses three text analysis modules, guided by a single domain discriminator for disentangled representation learning. Here, we suggest a novel optimization strategy that can enhance the quality of domain disentanglement, and also debilitates detrimental information of a source domain. Also, we extend the encoding network from a single to multiple domains, which has proven to be powerful for review-based recommender systems. Extensive experiments and ablation studies demonstrate that our method is efficient, robust, and scalable compared to the state-of-the-art single and cross-domain recommendation methods.