CNRS-Dataset-Atlas / README.md
PhysiQuanty's picture
Create README.md
27650d2 verified
|
Raw History Blame Contribute Delete
6.56 kB
metadata
license: apache-2.0
language:
  - fr
  - en
pretty_name: CNRS Dataset Atlas
tags:
  - open-science
  - scientific-data
  - research
  - dataverse
  - recherche-data-gouv
  - metadata
  - parquet
  - pyarrow

🔬 CNRS Dataset Atlas

A catalog of scientific metadata from Recherche Data Gouv, available in a machine learning-friendly format.

CNRS Dataset Atlas indexes 9,849 scientific datasets published or indexed on Recherche Data Gouv.

It brings together dataset descriptions, bibliographic references, authors, institutions, scientific disciplines, and persistent identifiers in a structured Apache Parquet format.

The catalog covers a wide range of scientific disciplines and is not limited to research conducted by the CNRS.

Dataset Overview

Property Value
Records 9,849
Columns 29
Format Apache Parquet
Structure One row per dataset
Source Recherche Data Gouv
Catalog License Apache-2.0
Organization CNRS-ML

The original scientific data files are not included. The catalog contains their metadata and references.

Dataset Structure

The dataset contains a single split, train, with 9,849 records.

Each record represents a scientific dataset identified by its title and persistent identifier.

Multivalued fields are represented as lists, while structured metadata is preserved as nested objects.

Data Fields

Identification

Field Type Description
name string Official dataset title.
type string Type of referenced resource.
url string Public URL of the dataset.
global_id string Persistent identifier, typically a DOI.
description string Scientific description provided by the authors.

Publication and Provenance

Field Type Description
published_at string Publication date returned by the API.
publisher string Publishing institution or repository.
identifier_of_dataverse string Dataverse collection identifier.
name_of_dataverse string Full name of the Dataverse collection.
storageIdentifier string Internal storage identifier of the resource.
createdAt string Resource creation date.
updatedAt string Last update date.

Bibliographic References

Field Type Description
citation string Complete bibliographic citation of the dataset.
citationHtml string Bibliographic citation in HTML format.
publications list Associated scientific publications, including references and URLs.
relatedMaterial list Related documentation and resources.

Scientific Classification

Field Type Description
subjects list Associated scientific disciplines.
keywords list Keywords describing the scientific content.
geographicCoverage list Geographic coverage, including country, region, city, and additional information.
dataSources list Data sources declared by the original producers.

Authors and Institutions

Field Type Description
authors list Names of scientific authors.
contacts list Scientific contacts and institutional affiliations.
producers list Institutions responsible for producing the data.

Files and Versions

Field Type Description
fileCount int64 Number of files associated with the dataset.
versionId int64 Dataverse version identifier.
versionState string Publication status of the dataset version.
majorVersion int64 Major version number.
minorVersion int64 Minor version number.

Fields may contain null values when the corresponding information is unavailable in the original source.

Scientific Coverage

The catalog covers a broad range of scientific disciplines, including:

  • Physics and Astronomy
  • Mathematics and Computer Science
  • Earth and Environmental Sciences
  • Biology and Life Sciences
  • Medicine and Health Sciences
  • Chemistry
  • Agricultural Sciences
  • Engineering
  • Social Sciences
  • Arts and Humanities

Descriptions and metadata are preserved in their original languages.

Data Example

Example of a catalog record:

{
  "name": "Replication Data for: Universal Relations in Scaling-Structured Turbulence",
  "type": "dataset",
  "url": "https://doi.org/10.57745/YJKPP7",
  "global_id": "doi:10.57745/YJKPP7",
  "publisher": "French Fluid Dynamics Database",
  "subjects": [
    "Physics"
  ],
  "keywords": [
    "Turbulence",
    "Hot-wire anemometry",
    "Fluid dynamics"
  ],
  "authors": [
    "Schmitt, Felix"
  ],
  "producers": [
    "Centre national de la recherche scientifique",
    "Université Grenoble Alpes"
  ],
  "fileCount": 7,
  "versionState": "RELEASED",
  "majorVersion": 2,
  "minorVersion": 0
}

Data Access

The catalog can be accessed directly using the Hugging Face Datasets library.

from datasets import load_dataset

dataset = load_dataset(
    "CNRS-ML/CNRS-Dataset-Atlas",
    split="train"
)

print(dataset)
print(dataset[0])

The Parquet files can also be explored using PyArrow, Pandas, or Polars.

Licensing

CNRS Dataset Atlas is released under the Apache-2.0 License.

This license applies to the published catalog. It does not replace the licenses or terms of use governing the original scientific datasets.

Each referenced resource retains its original scientific provenance and applicable reuse conditions.

Attribution

The metadata originates from Recherche Data Gouv.

Scientific works remain attributed to their respective authors and producing institutions.

To cite a specific dataset, refer to its bibliographic reference in the citation field and its persistent identifier in global_id.

CNRS-ML is an independent community initiative and is not an official CNRS publication.

Links