Data-driven analysis and visualization of dielectric properties curated from scientific literature
Tomoki Murata, Naoto Saito, Eiji Koyama, T. Phuong, Ryusuke Misawa, Satoshi Yokomizo, Tomoya Mato, Yu Takada, S. Hirose, Yukari Katsura
Starrydata From Published Graphs to Structured Research Data
Centered on 230,000+ experimental data points extracted from paper plots — complemented by crystal structure simulators, elemental reactivity maps, and other open tools for Materials Informatics research.
| Research Field | Papers | Figures | Samples | Curves |
|---|---|---|---|---|
| Loading… | ||||
What is Starrydata
Starrydata is an open data project that extracts and structures experimental data from graphs in inorganic materials science papers, making them freely reusable by researchers worldwide. Rather than a single project, it is a federation of research teams — each contributing curated data to a shared repository and together building a large open dataset that anyone can use.
How it works
Collect→Accumulate→Utilize
Experimental graphs from published papers are manually digitized and stored together with sample metadata in a single database. The resulting structured data is ready for visualization, download, and machine learning.
Data Collection
Data Utilization
Data Collection
Paper Selection
Identify papers, graphs, and plots as data collection targets
Graph Digitization
Trace curves in StarryDigitizer and convert graph data points to numerical values
Sample Descriptor Entry
Record composition, synthesis method, and measurement conditions from the paper
Open Database
Starrydata DB
Stores curves, sample metadata,
and paper information
and paper information
Data Utilization
Visualization & Exploration
Explore trends with Visualizer / Sample Explorer
Data Download
Freely download as CSV / JSON
ML & MI Research
Use structured data for model training and benchmarking
A Large-Scale Database
Patterns invisible in any single paper emerge when tens of thousands of samples and curves can be explored together.
Performance distributions across material families, composition–property correlations, outliers and unexplored composition space, and training / benchmark datasets for machine learning — these are the analyses that only a large-scale database makes possible.
Resources
Pick tools by purpose to work with Starrydata data.
What do you want to do?
Starrydata2 Web System
Register & search data
Register data from published plots, or search and browse existing samples, papers, and property curves
Datasets
Get datasets
Download CSV and per-field datasets
Visualizer
Visualize data
Explore composition and property spaces
Sample Explorer
Explore by composition
Filter samples by chemical composition
Resources
Explore all tools and documentation
Starrydata2 Web System (register, search, download data), plus visualization and composition-search tools (Visualizer / Sample Explorer / Explorer 3D), Figshare dataset snapshots, and the manual, data schema, and citation guide — organized by purpose.
Get Involved
Use the data, contribute data, partner on research, or support the project — there are many ways to get involved.
USE
Use the data
Everything you need to use Starrydata — web system, manual, Figshare snapshots and GitHub datasets, in one place.
Use the data →
COLLABORATE
Research & funding partnership
We welcome inquiries for joint research. Let's advance Materials Informatics together.
Contact us →Research
Seven research fields powering data curation and Materials Informatics studies.
Featured Publications
Top 3 of 8 project papers from the Starrydata team. 184 external publications cite our work.
Starrydata: from published plots to shared materials data
Yukari Katsura, Masaya Kumagai, Tomoya Mato, Yu Takada, Yuki Ando, Erina Fujita, Fumikazu Hosono, Eiji Koyama, Farhan Mudasar, T. Phuong, Naoto Saito, Yoshihiro Sakamoto, Atsumi Tanaka, Dewi Yana, Kaoru Kimura, Koji Tsuda, Masahiko Demura
Development of LLM-assisted data curation tools for the Starrydata materials science database
Yukari Katsura, Tomoya Mato, Yu Takada, Eiji Koyama, Dewi Yana, Atsumi Tanaka, Masaya Kumagai
Team
Starrydata runs as a dedicated team of researchers and engineers in materials science, data science, and web development.
Yukari Katsura
Project Leader
Masaya Kumagai
Web Development
Tomoya Mato
Web Development
Atsumi Tanaka
Project Operations (Research Administrator)
Dewi Yana
Data Analyst
Funders and Partners
NIMS
RIKEN AIP
JST
Univ. of Tsukuba
SAKURA internet