Agentic schema-guided extraction of materials process knowledge from scientific literature

2026-10-05

AI Summary

This study addresses the challenge of aggregating experimental knowledge from materials science literature, which is hindered by heterogeneous formats and context dependency. To this end, we propose SciKGExtract, a framework that integrates large language model-based extraction, PubChem chemical entity normalization, and multi-agent collaborative evaluation to enable structured extraction and knowledge graph integration of complex process knowledge. By introducing an agent-based iterative verification mechanism, the framework achieves an F1 score of 0.805 for ZnO extraction, significantly outperforming direct normalization baselines. Furthermore, this work effectively handles deeply nested schemas and elucidates the challenges inherent in multi-component processes. The results demonstrate the critical role of entity normalization and agent-based verification in generating machine-actionable knowledge from scientific literature.

Abstract

Materials literature contains detailed experimental knowledge, but procedures, chemical entities and measurements remain difficult to aggregate because they are reported in heterogeneous forms and depend on process-specific context. We present SciKGExtract, a schema-guided framework that combines large-language-model extraction with chemical normalization and agent-based evaluation and refinement before knowledge-graph integration. We evaluate the framework on 176 atomic-layer-deposition papers describing zinc oxide (ZnO) and indium--gallium--zinc oxide (IGZO), together with an expert-annotated full-schema subset. PubChem normalization improves exact-match extraction F1 for every tested model. For ZnO, the best F1 increases from 0.591 for direct normalized extraction to 0.805 with agentic refinement, whereas the best IGZO result is 0.344, revealing the greater difficulty of multicomponent supercycle processes. Evaluation against a deeply nested schema containing 65 experimental properties and 155 quantitative measurement nodes further exposes errors in process segmentation and numerical assignment. These results show that chemical canonicalization and targeted agentic verification provide complementary controls for converting complex materials literature into reusable, machine-actionable experimental knowledge.

Problem

• Materials process knowledge
• Knowledge extraction
• Scientific literature
• Chemical normalization
• Knowledge graph

Innovation

• Schema-guided extraction
• Large language models
• Agentic refinement
• Chemical normalization
• Knowledge graph