From Semantics to Hierarchy: A Hybrid Euclidean-Tangent-Hyperbolic Space Model for Temporal Knowledge Graph Reasoning
Siling Feng*, Zhisheng Qi* and Cong Lin
arXiv preprint arXiv:2409.00149v1 [cs.LG], 2024 (In submission to AAAI-25)
Temporal knowledge graphs (TKGs) have gained significant attention for their ability to extend traditional knowledge graphs with a temporal dimension, enabling dynamic representation of events over time. TKG reasoning involves extrapolation to predict future events based on historical graphs, which is challenging due to the complex semantic and hierarchical information embedded within such structured data. Existing Euclidean models capture semantic information effectively but struggle with hierarchical features. Conversely, hyperbolic models manage hierarchical features well but fail to represent complex semantics due to limitations in shallow models’ parameters and the absence of proper normalization in deep models relying on the L2 norm. Current solutions, such as curvature transformations, are insufficient to address these issues. In this work, a novel hybrid geometric space approach that leverages the strengths of both Euclidean and hyperbolic models is proposed. Our approach transitions from single-space to multi-space parameter modeling, effectively capturing both semantic and hierarchical information. Initially, complex semantics are captured through a fact co-occurrence and autoregressive method with normalizations in Euclidean space. The embeddings are then transformed into Tangent space using a scaling mechanism, preserving semantic information while relearning hierarchical structures through a query-candidate separated modeling approach, which are subsequently transformed into Hyperbolic space. Finally, a hybrid inductive bias for hierarchical and semantic learning is achieved by combining hyperbolic and Euclidean scoring functions through a learnable query-specific mixing coefficient, utilizing embeddings from hyperbolic and Euclidean spaces. Experimental results on four TKG benchmarks demonstrate that our method reduces error relatively by up to 15.0% in mean reciprocal rank (MRR) on YAGO compared to previous single-space models. Additionally, enriched visualization analysis validates the effectiveness of our approach, showing adaptive capabilities for datasets with varying levels of semantic and hierarchical complexity.