Contrastive Cross-domain Recommendation in Matching

Xie, Rong; Liu, Qi; Wang, Liangdong; Liu, Shukai; Zhang, Bo; Lin, Leyu

doi:10.1145/3534678.3539125

Cited by 55 publications

(8 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The basic idea is to extract valuable and transferable knowledge from many unlabeled data through a self-supervised task. Contrastive learning to solve problems such as cross-domain recommendation [33], social recommendation [34], and mitigating the long-tail effect of recommendation systems [35]. Xie et al [33] designed intra-domain comparisons and inter-domain comparisons to learn user representations and cross-domain knowledge transfer.…”

Section: Cotrastive Learningmentioning

confidence: 99%

“…Contrastive learning to solve problems such as cross-domain recommendation [33], social recommendation [34], and mitigating the long-tail effect of recommendation systems [35]. Xie et al [33] designed intra-domain comparisons and inter-domain comparisons to learn user representations and cross-domain knowledge transfer. We refer to cross-domain recommended methods.…”

Section: Cotrastive Learningmentioning

confidence: 99%

See 1 more Smart Citation

Heterogeneous Multi-Behavior Recommendation Based on Graph Convolutional Networks

et al. 2023

View full text Add to dashboard Cite

Traditional recommendation systems usually use single user-item interaction information, which ignore the multiple relationships that exist for other interactions (e.g., likes, clicks). Multi-behavioral recommendation models compensate for the shortcomings of traditional models. The existing multibehavior recommendation models focus on obtaining behavioral information by distinguishing the interaction differences of multiple user behaviors but ignore the common preferences of items related to user different interactions. In this paper, we propose a Multi-Behavior Heterogeneous Contrastive learning Recommendation (MBHCR) model. Specifically, MBHCR stresses the information fusion between different interaction behaviors and differences in the behavior view. Firstly, we design a unified heterogeneous graph based on the complexity of user multi-behavioral interaction information to distinguish behavioral differences while preserving their complete semantic information. Secondly, we propose a multi-behavior relational aggregator component to model the unified heterogeneous graph only once to capture the potential common representations of users and their different interactions and to mitigate user sparsity. Thirdly, we also design a behavior comparison learning enhancer to complement the interaction differences between the user's target behavior and auxiliary behaviors on the behavior view and extract valid information. Experiments were conducted on two real-world datasets to demonstrate the validity of MBHCR and designed ablation experiments to verify the contribution of model components.

show abstract

Section: Cotrastive Learningmentioning

confidence: 99%

Section: Cotrastive Learningmentioning

confidence: 99%

Heterogeneous Multi-Behavior Recommendation Based on Graph Convolutional Networks

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The Self-Supervised Learning (SSL) paradigm has been widely studied in multiple research communities, e.g., Computer Vision [3,5,12,15,33,34], Natural Language Processing [7,23], and non-sequential recommendation systems [2,51,56,[62][63][64]. SSL has recently been introduced into sequential recommendation systems to alleviate the data sparsity issue.…”

Section: Self-supervised Learning In Srmentioning

confidence: 99%

APGL4SR: A Generic Framework with Adaptive and Personalized Global Collaborative Information in Sequential Recommendation

Yin,

Wang,

et al. 2023

Proceedings of the 32nd ACM International Conference on Information and Knowledge Management

View full text Add to dashboard Cite

The sequential recommendation system has been widely studied for its promising effectiveness in capturing dynamic preferences buried in users' sequential behaviors. Despite the considerable achievements, existing methods usually focus on intra-sequence modeling while overlooking exploiting global collaborative information by inter-sequence modeling, resulting in inferior recommendation performance. Therefore, previous works attempt to tackle this problem with a global collaborative item graph constructed by pre-defined rules. However, these methods neglect two crucial properties when capturing global collaborative information, i.e., adaptiveness and personalization, yielding sub-optimal user representations. To this end, we propose a graph-driven framework,

show abstract

“…Given word embedding 𝑅 H 𝑢 and 𝑅 H 𝑖 from each domain, we aim to extract useful information. Several feature extraction strategies have been proposed, including domain adaptation [71], variational reconstruction [41], personalized transfer [78], contrastive learning [64], and domain disentanglement [49]. Among them, we adopt the domain disentanglement algorithm, which does not obligate user overlapping in both domains [8].…”

Section: Feature Extractionmentioning

confidence: 99%

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

Choi¹,

Choi²,

Ko³

et al. 2022

Proceedings of the 31st ACM International Conference on Information &Amp; Knowledge Management

View full text Add to dashboard Cite

The issue of data sparsity poses a significant challenge to recommender systems. In response to this, algorithms that leverage side information such as review texts have been proposed. Furthermore, Cross-Domain Recommendation (CDR), which captures domainshareable knowledge and transfers it from a richer domain (source) to a sparser one (target), has received notable attention. Nevertheless, the majority of existing methodologies assume a Euclidean embedding space, encountering difficulties in accurately representing richer text information and managing complex interactions between users and items. This paper advocates a hyperbolic CDR approach based on review texts for modeling user-item relationships. We first emphasize that conventional distance-based domain alignment techniques may cause problems because small modifications in hyperbolic geometry result in magnified perturbations, ultimately leading to the collapse of hierarchical structures. To address this challenge, we propose hierarchy-aware embedding and domain alignment schemes that adjust the scale to extract domainshareable information without disrupting structural forms. The process involves the initial embedding of review texts in hyperbolic space, followed by feature extraction incorporating degree-based normalization and structure alignment. We conducted extensive experiments to substantiate the efficiency, robustness, and scalability of our proposed model in comparison to state-of-the-art baselines. CCS CONCEPTS• Information systems → Recommender systems.

show abstract

Contrastive Cross-domain Recommendation in Matching

Cited by 55 publications

References 34 publications

Heterogeneous Multi-Behavior Recommendation Based on Graph Convolutional Networks

Heterogeneous Multi-Behavior Recommendation Based on Graph Convolutional Networks

APGL4SR: A Generic Framework with Adaptive and Personalized Global Collaborative Information in Sequential Recommendation

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

Contact Info

Product

Resources

About