Skip to content

ComProScanner Published in Digital Discovery

ComProScanner TOC Graphic

Overview

ComProScanner has been published in Digital Discovery, a journal of the Royal Society of Chemistry. The paper, titled "ComProScanner: a multi-agent based framework for composition-property structured data extraction from scientific literature", presents an autonomous multi-agent platform for extracting, validating, and visualising machine-readable chemical compositions and properties from journal articles, creating structured, ML-ready databases with no human intervention.

1. The Research

ComProScanner establishes a scalable, user-friendly pathway for building large experimental datasets from journal articles, directly addressing one of the core bottlenecks in data-driven materials discovery. By combining LLM agents with RAG, domain-specific embeddings, and publisher TDM APIs, it bridges the gap between unstructured scientific literature and machine learning pipelines.

Key Features

  • Publisher-to-database pipeline via Text and Data Mining (TDM) APIs from Elsevier, Springer Nature, IOP Publishing, and Wiley — no manual PDF uploads required
  • Five specialised AI agents organised into a RAG-based filtering crew, a composition extraction crew, and a synthesis information crew
  • Variable composition resolution using a deep learning tool to expand formulas like Na(1−x)LixTiO₃ into distinct compounds
  • Built-in evaluation framework with weight-based accuracy, classification metrics, and novel normalised classification metrics
  • Implementable in fewer than 20 lines of Python code via pip install comproscanner

2. Results

The framework was benchmarked on 100 journal articles across 10 LLMs for extracting piezoelectric material compositions and d33 coefficient values. DeepSeek-V3-0324 achieved the best overall accuracy of 0.82, with a compositional accuracy of 0.90. Strikingly, over 99% of the extracted piezoelectric materials are absent from the Materials Project database, highlighting the untapped wealth of data buried in the literature.

3. About the Authors

This work was led by Aritra Roy, a PhD student in the Energy, Materials and Environment Research Centre at LSBU, under the supervision of Prof. John Buckeridge (LSBU) and Dr. Chiara Gattinoni (King's College London), with contributions from Prof. Enrico Grisan (LSBU). Financial and legal support for publisher TDM licences was provided by LSBU and KCL. Dr. Gattinoni's work was supported by the EPSRC New Investigator Award [grant number UKRI132].

4. Publication Details

Title: ComProScanner: a multi-agent based framework for composition-property structured data extraction from scientific literature
Journal: Digital Discovery
Authors: Aritra Roy, Enrico Grisan, John Buckeridge, and Chiara Gattinoni
DOI: 10.1039/D5DD00521C
Link: Read the full paper here

Citation (BibTeX):

@Article{roy2026comproscannermultiagentbasedframework,
  author    = {Roy, Aritra and Grisan, Enrico and Buckeridge, John and Gattinoni, Chiara},
  title     = {ComProScanner: a multi-agent based framework for composition-property structured data extraction from scientific literature},
  journal   = {Digital Discovery},
  year      = {2026},
  pages     = {Accepted},
  publisher = {RSC},
  doi       = {10.1039/D5DD00521C},
  url       = {https://doi.org/10.1039/D5DD00521C}
}

5. Access ComProScanner

6. Looking Ahead

ComProScanner is fully open source and available for the community to use, extend, and adapt to new properties and material systems. Future directions include integrating OCR or vision-language models (VLMs) to extract data from figures and graphs, and extending the framework to support extraction of multiple material properties simultaneously.