DISCIE-Discriminative Closed Information Extraction

2025-06-19

International Workshop on the Semantic Web

AI Summary

To address the accuracy bottleneck in large-scale closed-domain information extraction—caused by millions of entities and hundreds of relations (especially long-tail ones)—this paper proposes a discriminative relation extraction framework. The method introduces type-aware embeddings that jointly encode entity types and relation-specific semantics, and designs an entity-relation joint scoring mechanism, enabling end-to-end discriminative modeling within a lightweight neural architecture. Compared to state-of-the-art generative approaches, our method achieves significantly higher F1 scores for both overall and long-tail relations while maintaining superior inference efficiency: average F1 improves by 4.2% and long-tail relation F1 by 9.7% on mainstream closed-domain benchmarks. Our core contribution is the first systematic integration of type-guided discriminative modeling into large-scale closed-domain extraction, achieving accuracy and generalization surpassing large language models with substantially fewer parameters.

Technology Category

• Natural Language Processing: Information Extraction
• Reasoning under Uncertainty: Relational Probabilistic Models
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This paper introduces a novel method for closed information extraction. The method employs a discriminative approach that incorporates type and entity-specific information to improve relation extraction accuracy, particularly benefiting long-tail relations. Notably, this method demonstrates superior performance compared to state-of-the-art end-to-end generative models. This is especially evident for the problem of large-scale closed information extraction where we are confronted with millions of entities and hundreds of relations. Furthermore, we emphasize the efficiency aspect by leveraging smaller models. In particular, the integration of type-information proves instrumental in achieving performance levels on par with or surpassing those of a larger generative model. This advancement holds promise for more accurate and efficient information extraction techniques.

Problem

• Improves relation extraction accuracy using discriminative approach
• Addresses large-scale closed information extraction challenges
• Enhances efficiency with smaller models and type-information

Innovation

• Discriminative approach with type and entity info
• Efficient smaller models outperform larger ones
• Superior accuracy for long-tail relations