Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 27, 2025

Temporal knowledge graph reasoning(TKGR) has attracted widespread attention due to its ability handle dynamic temporal features. However, existing methods face three major challenges: (1) the difficulty of capturing long-distance dependencies in information sparse environments; (2) problem noise interference; (3) complexity modeling relationships. These seriously impact accuracy and robustness reasoning. To address these challenges, we proposes a framework based on Dual-gate Noise-aware Contrastive Learning (DNCL) improve performance TKGR. The consists core modules: We employ multi-dimensional gated update module, which flexibly selects key suppresses redundant through dual-gate mechanism, thereby alleviating problem; construct noise-aware adversarial improves reduces training; design multi-layer embedding contrastive learning enhances representation intra-layer inter-layer strategies better capture latent relationships dimension. Experimental results four benchmark datasets show that DNCL model is than current methods, especially for ICEWS14, ICEWS05-15 ICEWS18 datasets, Hit@1 improved by 6.91%, 4.31% 5.30% respectively.

Язык: Английский