A Unified Temporal Knowledge Graph Reasoning Model Towards Interpolation and Extrapolation

TL;DR

This work presents TPAR, a unified neural-symbolic model for temporal knowledge graph reasoning that handles both interpolation and extrapolation. By combining a Bellman-Ford-inspired recursive encoding of temporal paths with neural scoring and symbolic path reasoning, TPAR achieves robust, interpretable predictions under noisy temporal data. The approach introduces relative time encoding and a recursive path-encoding mechanism to score destination entities, enabling explicit evidence paths for decisions. Empirical results across multiple interpolation and extrapolation datasets show that TPAR outperforms state-of-the-art baselines, and pipeline analyses demonstrate the model’s ability to integrate both settings in a unified framework with interpretable reasoning traces.

Abstract

Temporal knowledge graph (TKG) reasoning has two settings: interpolation reasoning and extrapolation reasoning. Both of them draw plenty of research interest and have great significance. Methods of the former de-emphasize the temporal correlations among facts sequences, while methods of the latter require strict chronological order of knowledge and ignore inferring clues provided by missing facts of the past. These limit the practicability of TKG applications as almost all of the existing TKG reasoning methods are designed specifically to address either one setting. To this end, this paper proposes an original Temporal PAth-based Reasoning (TPAR) model for both the interpolation and extrapolation reasoning. TPAR performs a neural-driven symbolic reasoning fashion that is robust to ambiguous and noisy temporal data and with fine interpretability as well. Comprehensive experiments show that TPAR outperforms SOTA methods on the link prediction task for both the interpolation and the extrapolation settings. A novel pipeline experimental setting is designed to evaluate the performances of SOTA combinations and the proposed TPAR towards interpolation and extrapolation reasoning. More diverse experiments are conducted to show the robustness and interpretability of TPAR.

Figures (5)

• Figure 1: An illustration of Bellman-Ford-based recursive encoding.
• Figure 2: Path interpretations of TKG reasoning on ICEWS14 test set. Dashed lines mean inverse relations.
• Figure 3: Performance w.r.t. sparsity ratio on ICEWS14.
• Figure 4: An illustration of temporal paths and temporal links.
• Figure 5: The MRR performance with different maximum path lengths for the extrapolation reasoning on the ICEWS14 test set.

Theorems & Definitions (2)

• Definition 1
• Definition 2