Beyond Pairwise: Global Zero-shot Temporal Graph Generation

2025-02-16

AI Summary

Existing temporal relation extraction (TRE) methods rely on pairwise event classification, suffering from low efficiency and difficulty in ensuring global temporal consistency. This paper proposes the first zero-shot, end-to-end temporal graph generation framework for full-document TRE: it abandons the pairwise discrimination paradigm and directly generates a complete temporal graph over all events; incorporates logical rule-based transitivity constraints and consistency regularization to explicitly model temporal logic; and introduces OmniTemp—the first fully pairwise annotated dataset for document-level TRE. Experiments demonstrate that our method significantly outperforms prior zero-shot TRE models, achieving over 12% absolute F1 improvement, while matching the performance of supervised approaches. Our work establishes a novel paradigm for document-level temporal structure modeling.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Temporal Planning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Temporal relation extraction (TRE) is a fundamental task in natural language processing (NLP) that involves identifying the temporal relationships between events in a document. Despite the advances in large language models (LLMs), their application to TRE remains limited. Most existing approaches rely on pairwise classification, in which event pairs are considered individually, leading to computational inefficiency and a lack of global consistency in the resulting temporal graph. In this work, we propose a novel zero-shot method for TRE that generates a document's complete temporal graph at once, then applies transitive constraints optimization to refine predictions and enforce temporal consistency across relations. Additionally, we introduce OmniTemp, a new dataset with complete annotations for all pairs of targeted events within a document. Through experiments and analyses, we demonstrate that our method significantly outperforms existing zero-shot approaches while achieving competitive performance with supervised models.

Problem

• Enhances temporal relation extraction efficiency
• Generates complete temporal graphs globally
• Introduces new dataset for comprehensive evaluation

Innovation

• Zero-shot temporal graph generation
• Transitive constraints optimization
• OmniTemp dataset for complete annotations