DOREMI: Optimizing Long Tail Predictions in Document-Level Relation Extraction

2026-01-01

Knowledge-Based Systems

AI Summary

This work addresses the challenge of long-tailed relation distributions in document-level relation extraction (DocRE), where scarce samples of rare relations hinder model generalization. To mitigate this issue, the authors propose a lightweight iterative active learning framework that selectively acquires the most informative samples for annotation with minimal labeling cost, without relying on large-scale noisy data or complex denoising mechanisms. The framework is designed to seamlessly integrate with any existing DocRE model, effectively alleviating long-tail bias. By iteratively refining the model through strategically chosen examples, the approach not only enhances performance on rare relations but also improves overall training efficiency.

Technology Category

• Machine Learning: Active Learning
• Natural Language Processing: Information Extraction
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Problem

• Document-Level Relation Extraction
• long-tail distribution
• relation types
• data scarcity
• underrepresented relations

Innovation

• Document-Level Relation Extraction
• Long-Tail Optimization
• Active Learning
• Iterative Framework
• Data Efficiency