Reliable End-to-End Material Information Extraction from the Literature with Source-Tracked Multi-Stage Large Language Models

2025-10-01

AI Summary

Existing material literature information extraction methods suffer from narrow feature coverage and difficulty in modeling multidimensional correlations among composition, processing, microstructure, and properties (C–P–M–P). To address this, we propose a large language model–based multi-stage source-tracing extraction framework. It performs iterative entity recognition, disambiguation, relation inference, and provenance verification to end-to-end extract comprehensive experimental data spanning 47 fine-grained features. Our approach achieves the first systematic joint modeling and traceable extraction of the full C–P–M–P chain. It attains F1 scores of 0.96 at both feature-level and tuple-level evaluation; microstructure-related F1 improves by 10.0–13.7%; the number of missed materials drops from 49 to 13; and zero false positives are observed. The framework significantly enhances the completeness, accuracy, and interpretability of materials databases.

Technology Category

• Natural Language Processing: Information Extraction
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Data-driven materials discovery requires large-scale experimental datasets, yet most of the information remains trapped in unstructured literature. Existing extraction efforts often focus on a limited set of features and have not addressed the integrated composition-processing-microstructure-property relationships essential for understanding materials behavior, thereby posing challenges for building comprehensive databases. To address this gap, we propose a multi-stage information extraction pipeline powered by large language models, which captures 47 features spanning composition, processing, microstructure, and properties exclusively from experimentally reported materials. The pipeline integrates iterative extraction with source tracking to enhance both accuracy and reliability. Evaluations at the feature level (independent attributes) and tuple level (interdependent features) yielded F1 scores around 0.96. Compared with single-pass extraction without source tracking, our approach improved F1 scores of microstructure category by 10.0% (feature level) and 13.7% (tuple level), and reduced missed materials from 49 to 13 out of 396 materials in 100 articles on precipitate-containing multi-principal element alloys (miss rate reduced from 12.4% to 3.3%). The pipeline enables scalable and efficient literature mining, producing databases with high precision, minimal omissions, and zero false positives. These datasets provide trustworthy inputs for machine learning and materials informatics, while the modular design generalizes to diverse material classes, enabling comprehensive materials information extraction.

Problem

• Extracting comprehensive materials data trapped in unstructured scientific literature
• Capturing integrated composition-processing-microstructure-property relationships from experiments
• Building reliable databases with minimal omissions and zero false positives

Innovation

• Multi-stage LLM pipeline extracts 47 material features
• Integrates source tracking for enhanced accuracy
• Modular design generalizes across diverse material classes