跳转至

CoRCi：跨域序列推荐中相干兴趣的交叉重构建模

文章背景与核心概要

跨域序列推荐（CDSR）旨在通过在相关领域间迁移动态用户兴趣来缓解数据稀疏性问题。然而，现有的方法在平衡领域特定兴趣与领域不变兴趣时往往面临挑战，容易放大领域间的差异，从而破坏兴趣的相干性。

为了解决这些局限性，本文提出了 CoRCi（Coherent Interest Cross-Reconstruction，相干兴趣交叉重构）框架。该框架通过交叉注意力机制直接从预编码的特定领域表示中生成混合领域表示，并采用单一的序列级领域无关损失函数进行优化。此外，CoRCi 引入了 FocalNCE 损失函数，通过对同领域负样本施加更高惩罚，有效强化了领域不变性的对齐。在四个真实数据集上的实验表明，CoRCi 在各项指标上均显著优于当前最先进的 CDSR 方法。

📋 摘要 (Summary)

Cross-Domain Sequential Recommendation (CDSR) transfers dynamic user interests between related domains to mitigate data sparsity. However, existing methods struggle to balance domain-specific and domain-invariant interests, often magnifying inter-domain discrepancies and disrupting interest coherence.

To overcome these limitations, the authors introduce CoRCi (Cross-Reconstruction for Coherent Interest), a dual-target CDSR framework featuring: 1. Cross-Reconstruction: Generates mixed-domain representations directly from pre-encoded specific-domain representations using cross-attention, optimized via a single sequence-level, domain-agnostic loss. 2. FocalNCE: Integrates Focal Loss into the mixed-domain InfoNCE objective to penalize negative samples from the same domain as the query, strengthening domain-invariant alignment.

Extensive experiments on four real-world datasets show that CoRCi consistently outperforms current state-of-the-art CDSR methods across all evaluation metrics.

👥 作者 (Authors)

  • Qingtian Bian
  • Tieying Li
  • Marcus de Carvalho
  • Jiaxing Xu
  • Hui Fang
  • Yiping Ke

📄 论文摘要 (Abstract)

跨域序列推荐（CDSR）旨在通过在相关领域间迁移动态用户兴趣来缓解数据稀疏性问题。有效连接这些领域是该领域的核心挑战。在单域建模中，模型无法区分领域特定兴趣和领域不变兴趣。近期的方法通常将领域特定序列按时间顺序合并为混合领域序列，以捕获领域不变知识。然而，这些方法通常为混合领域序列部署独立的编码器，并使用各领域损失聚合的方式进行训练。这种工作流放大了领域间的差异，并破坏了领域不变兴趣的相干性，特别是在序列到序列（Seq2Seq）任务中查询目标对源自不同领域时。在本文中，我们提出了 CoRCi（相干兴趣交叉重构），这是一个旨在解决上述缺陷的双目标 CDSR 框架。具体而言，CoRCi 提出了一种交叉重构方法，通过交叉注意力机制直接从预编码的特定领域表示中生成混合领域表示。生成的表示随后使用单一的、序列级的、领域无关的损失进行训练，以保持领域不变兴趣的相干性。为了进一步抑制混合领域建模中的领域差异，CoRCi 引入了 FocalNCE，将 Focal Loss 嵌入到前述的混合领域 InfoNCE 目标中。新的损失函数为与查询同领域的负样本分配更高的惩罚，从而加强了领域不变性的对齐。在四个真实数据集上的广泛实验表明，CoRCi 在所有指标上均持续优于当前最先进的 CDSR 方法，并取得了统计学意义上的显著提升。

Cross-Domain Sequential Recommendation (CDSR) aims to alleviate data sparsity by transferring dynamic user interests across related domains. A key challenge lies in effectively bridging these domains. In single-domain modeling, models cannot distinguish between domain-specific and domain-invariant interests. Recent methods merge domain-specific sequences chronologically into a mixed-domain sequence to capture domain-invariant knowledge. However, they typically deploy separate encoders for the mixed-domain sequence and train them with per-domain loss aggregation. This workflow magnifies inter-domain discrepancies and disrupts domain-invariant interest coherence, especially when query target pairs in Seq2Seq originate from different domains. In this paper, we present CoRCi (Cross-Reconstruction for Coherent Interest), a dual-target CDSR framework that tackles these drawbacks. Specifically, CoRCi proposes a Cross-Reconstruction approach that generates mixed-domain representations directly from pre-encoded specific-domain representations via cross-attention. The generated representations are then trained using a single, sequence-level, domain-agnostic loss to preserve the coherence of domain-invariant interests. To further suppress domain discrepancies in mixed-domain modeling, CoRCi introduces FocalNCE, which embeds Focal Loss into the preceding mixed-domain InfoNCE objective. The new loss assigns higher penalties to negatives drawn from the same domain as the query, thereby strengthening domain-invariant alignment. Extensive experiments on four real-world datasets demonstrate that CoRCi consistently outperforms state-of-the-art CDSR counterparts, achieving statistically significant gains across all metrics.

license icon