ComProScanner: A multi-agent based framework for composition-property structured data extraction from scientific literature

2025-10-23

AI Summary

To address the challenges of low-efficiency, poor verifiability, and difficulty in structured extraction of ceramic piezoelectric material data—particularly chemical compositions and the piezoelectric strain coefficient *d*₃₃—from scientific literature, this work proposes a multi-agent collaborative framework. It integrates open- and closed-source large language models (LLMs), rule-based engines, and consistency verification mechanisms to enable end-to-end automated extraction, validation, classification, and visualization of composition–property data. Evaluated on 100 peer-reviewed journal articles—with DeepSeek-V3-0324 as the primary LLM—the framework achieves an overall accuracy of 82% and constructs the first high-quality, reusable, machine-learning-ready structured database for piezoelectric ceramics. Its core innovation lies in pioneering the application of multi-agent collaboration to materials literature data engineering, uniquely balancing deep semantic understanding with rigorous physical constraints—thereby significantly improving both the accuracy and reproducibility of automated mining of complex experimental data.

Technology Category

• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Natural Language Processing: Information Extraction
• Multiagent Systems: Coordination and Collaboration

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Since the advent of various pre-trained large language models, extracting structured knowledge from scientific text has experienced a revolutionary change compared with traditional machine learning or natural language processing techniques. Despite these advances, accessible automated tools that allow users to construct, validate, and visualise datasets from scientific literature extraction remain scarce. We therefore developed ComProScanner, an autonomous multi-agent platform that facilitates the extraction, validation, classification, and visualisation of machine-readable chemical compositions and properties, integrated with synthesis data from journal articles for comprehensive database creation. We evaluated our framework using 100 journal articles against 10 different LLMs, including both open-source and proprietary models, to extract highly complex compositions associated with ceramic piezoelectric materials and corresponding piezoelectric strain coefficients (d33), motivated by the lack of a large dataset for such materials. DeepSeek-V3-0324 outperformed all models with a significant overall accuracy of 0.82. This framework provides a simple, user-friendly, readily-usable package for extracting highly complex experimental data buried in the literature to build machine learning or deep learning datasets.

Problem

• Extracts structured composition-property data from scientific literature
• Automates validation and visualization of experimental material datasets
• Addresses scarcity of tools for building machine learning-ready databases

Innovation

• Multi-agent platform for structured data extraction
• Autonomous system integrating validation and visualization
• Framework tested across multiple LLMs for accuracy