RO-Crate RDA maDMP Mapper

Ro-Crate RDA maDMP Mapper is a tool that defines mapping between RO-Crate and RDA maDMP. It allows generating RO-Crate(s) from maDMP (1 to many) and maDMP from RO-Crate(s) (many to 1).

Mapping

Assumptions

DMPs can include multiple datasets. However, ro-crate include only one dataset. The dataset can include other nested datasets. It is assumed that most of the data for the dataset at the root (e.g. funding, contributors, authors,...) is equivalent to the data of DMP. Nested datasets will be included as elements of distributions.

RO-crates have to have @id for each property. Otherwise, the jsonld generated would have a wrong format. However, an identifier is not always present for entities of DMP. In case of absence of identifiers, it is assumed that the title is also the @id.

The website of ro-crate explicitly mentions that it is possible to use schema.org metadata to supplement RO-Crate. So, if there are attributes which are present in DMP and missing in ro-crates, they are accounted for by other Linked Data Vocabularies. (Example: cost)

Mapped Properties

ro-crate attribute	@type of ro-crate property	DMP attribute	Parent of DMP attribute	Comment/Assumption
contactPoint	ContactPoint	contact	dmp	It is assumed that the contact person for the DMP is the same as the contact person of the dataset.
@id	ContactPoint	identifier	contact_id
email	ContactPoint	mbox	contact
name	ContactPoint	name	contact
contactType	ContactPoint	type	contact	These attributes are not exactly equivalent. But, they are close enough.
author/creator	Dataset	contributor	dmp
@id	person	identifier	contributor_id
affiliation	person	role	contributor
name	person	name	contributor
email	person	mbox	contributor
@id	person	identifier	contributor_id
cost	Dataset	cost	dmp	In DMP, the cost represents a list of costs related to data management. However, the cost for ro-crate may not include all costs.
costCurrency	cost	currency_code	cost	This is not explicitly mentioned in ro-crate website. But, cost properties can be found in jsonld context used for ro-crates.
description	cost	description	cost
@id	cost	title	cost
value	cost	value	cost
Language	Dataset	language	dmp
description	Dataset	description	dataset
name	Dataset	title	dataset
identifier	Dataset	identifier	dataset_id
Dataset	hasPart	distribution	dataset	Ro-crates usually include a lot of sub-datasets. They usually include information about the main dataset like encoding and size. Therefore, they are accounted for as distributions.
File	hasPart	distribution	dataset	Ro-crates usually include a lot of sub-Files. Sub-datasets have sub-Files. Deeply nested files will not be taken into consideration. This is because it does not aline with the concept of distribution.
downloadUrl	DataDownload	download_url	distribution
contentUrl	DataDownload	access_url	distribution
endDate	DataDownload	available_until	distribution
contentSize	Dataset	byte_size	distribution
contentSize	Dataset/File	byte_size	distribution
encodingFormat	Dataset/File	format	distribution
name	contentLocation	geo_location	host	It is assumed that the name is a country. But, it is not always the case. But, it is better than losing th e information.
title	RepositoryObject	title	host
description	RepositoryObject	description	host
@id	RepositoryObject	url	host
availability	RepositoryObject	availability	host	availability is not defined exactly the same for ro-crate and DMP in terms of the format of the inputed value.
license	Dataset/File	license	distribution
@id	license	license_ref	license
identifier	license	license_ref	license
datePublished	Dataset	issued	dataset
keywords	Dataset	keyword	dataset
description	Organisation	description	project
name	Organisation	title	project
name	Organisation	title	project
endDate	Organisation	end	project
startDate	Organisation	start	project
@id	Grant	identifier	funder_id
@id	CreativeWork	identifier	metadata_standard_id
Language	CreativeWork	language	dmp	metadata
description	CreativeWork	description	dmp	metadata

Unmapped Properties of DMP

DMP attribute	Comment
dmp	There is no direct mapping to dmp since ro-crate is an approach to package research data with their metadata. DMP concept considers the bigger picture.
contact_id, contributor_id, dmp_id, dataset_id, hasPart, host, project, metadata_standard_id, funder_id	For these attributes, there isn't an equivalent attribute. But, their children have equivalent attributes. The equivalence is not needed and it is not classified as missing.
type	"type" of "contact_id" and "contributor_id" have different meaning than "@type". There isn't an equivalent.
created	The DMP creation should be at the time the DMP is generated. Therefore, it does not have an equivalent in ro-crates.
modified	DMP creation date is the same as the modified date in this case.
personal_data, preservation_statement, preservation_statement, security_and_privacy, data_access, storage_type, pid_system, certified_with, backup_type, backup__frequency, ethical_issues_description, ethical_issues_exist, ethical_issues_report, data_quality_assurance, sensitive_data	Almost all attributes which have to do with quality, privacy, ethics and security are missing in ro-crate and connot be translated. PS: accessMode in https://w3id.org/ro/crate/1.0/context is different from data_access.
support_versioning	This specific attribute has no equivalent in the context. However, there is the possibility to include the specific versions with other attributes.
funding_status	The definition provided in DMP official description does not match any of the descriptions for the ro-crate.
identifier, description, title	These attributes have to be inputed by the user. It is wrong to assume that the DMP's identifier is the same as the dataset's identifier. The user should create a new identifier for the DMP. It is also not always logical to give the DMP the name of the dataset.

Unmapped Properties of ro-crate

ro-crate attribute	Comment
@context, @graph, conformsTo, about	There isn't an equivalent for these attributes because they are specific to jsonld. However, DMP does not follow jsonld schema. It follows json schema. When automatically generating a ro-crate, they are inserted as they can be programatically determined. When generating DMPs, they are ignored.
hasPart, hasMember	DMPs do not have the equivalent of these attributes. However, they are accounted for by including their values in distributions.
publisher, sameAs, temporalCoverage	These attributes are not included as explicitly in DMPs.
CreateAction, UpdateAction	CreateAction and UpdateAction classes are there to model the contributions of Context Entities of type Person or Organization. This is not present in DMPs.
latitude, longitude	Places are described more thouroughly in ro-crates.

Usage

In order to map between both standards, the following command should be used:

python mapper.py -i <input_path> -o <output_path>

Where input_path represents a path to the input folder (or file) representing an maDMP or a RO-Crate project and output_path represents a path where the generated files will be stored.

It is not required to explicitally define the mapping direction (maDMP to RO-Crate or RO-Crate to maDMP) since this is handled within the tool. x

Examples

Several maDMP and RO-Crate examples are also provided. They are structured in the following way:

├── madmp
│   ├── calculation-of-nice-sunny-days
│   ├── closed
│   ├── dataset-many
│   ├── funded-project
│   ├── life-expectancy-prediction
│   ├── long
│   ├── minimal-content
│   ├── multilayer-perceptron-on-hypothyroid
│   ├── swedish-motor-insurance
│   └── world-development-indicators
└── rocrate
    ├── drug_consumption
    ├── Glop_Pot
    ├── GTM
    ├── NursingResidentStuff
    └── world_development_indicators_visualization

For maDMPs, 5 of the examples are taken from examples made by students as part of Data Stewardship at Vienna University of Technology. The other 5 are taken from the examples provided by the RDA-DMP-Common-Standard.

For RO-Crates, 3 examples are taken from the official RO-Crate website and the other 2 are especially created to test the coverage of the mapping using existing datasets.

Demo

A demo of the mapping between both standards can be executed using the following command:

python demo.py

This demo uses the content examples folder in order to generate mapped files and store them in demo_files.

Running tests

In order to run the unit tests stored under tests, pytest can be used. To do so, navigate to tests folder and run the following command pytest -vv. Unit tests are made to test the functionality of some methods within the different modules and are not testing the mapping functionality.

Contributors

Ghaith Arfaoui

Maroua Jaoua

License

MIT License