Starrydata: from published plots to shared materials data

doi:10.1080/27660400.2025.2506976

Starrydata: from published plots to shared materials data

2025-12-31

8

PRE

AI

DOI：10.1080/27660400.2025.2506976

Original

Original request for help

Save

Abstract

En 中文

We have developed the Starrydata2 web system, an open, web-based database for collecting and organizing experimental material property data from the literature. It assists users worldwide in extracting and sharing curve data from plot images in published papers, along with relevant sample information such as chemical compositions and fabrication methods. Starrydata2 streamlines the manual data collection process through partial automation. Currently, Starrydata encompasses over 194,000 curves extracted from more than 82,000 physical samples, as reported in over 13,000 publications on functional inorganic materials, including thermoelectric and magnetic materials. All data in Starrydata are openly accessible to the public for both commercial and non-commercial purposes. In this paper, we introduce the web interface, data curation workflow, data structure, and system architecture of Starrydata2. We then described in detail the datasets currently included in Starrydata2 and discuss their use cases. We also present the methods for applying the collected dataset, including a unique large-scale data representation method called 'all-data plots', which provides a comprehensive overview of the entire dataset. Finally, we report on how the collected datasets are being utilized in data-driven materials research through machine learning, modelling and simulation.

Keywords:

Materials informatics

database

data mining

plot digitization

open data

literature data

inorganic materials

functional materials

thermoelectric materials

magnetic materials

Starrydata: from published plots to shared materials data

Abstract

Journal

Organization