Schema 2.0
Community feedback
In the spring of 2023, ROR asked for community feedback on the specifics of major, breaking changes to the ROR metadata schema.
- ROR Schema v2.0 draft proposal - proposal open for comment through February 2023
- ROR Schema v2.0 second draft proposal - proposal open for comment through April 2023
- ROR Schema v2.0 final proposal - proposal adopted April 2023
Planned changes
Work to implement the below agreed-upon changes to the current ROR metadata schema will continue throughout 2023 , and version 2.0 of the ROR metadata schema is expected to be released in the last quarter of 2023 or early in 2024. The current version (unofficially v1.0) will be maintained for at least a year after the release of v2.0.
Read more about how ROR will handle Schema versions.
Name information (name, acronyms, aliases, labels)
ROR records in version 1 currently include 4 separate fields that represent variations of an organization’s name: name
, aliases
, labels
and acronyms
. User-reported problems with these fields include
- the lack of language code in the
name
field, which makes it difficult to utilize ROR data in internationalized applications or in cases where names need to be displayed in a single preferred language, and - issues with the concept of a "primary" name that can have only one value, since in multilingual countries such as Canada and Switzerland it is important that names in different languages be given equal weight.
We are therefore "flattening" all name-related fields into one names
field that can hold an array of values, types, and languages. For use cases where a single "default" name for an organization is desired, we have added the ror_display
type to indicate that this is the name that ROR has chosen to display in its web-based search.
Current v1.0 example
"acronyms" : [
"UC"
],
"aliases" : ["UC System"],
"labels" : [
{
"iso639" : "es",
"label" : "Universidad de California"
},
{
"iso639" : "fr",
"label" : "Université de Californie"
}
],
"name" : "University of California System",
Forthcoming v2.0 example
"names" : [
{
"value" : "UC",
"types": ["acronym"],
"lang" : "en"
},
{
"value" : "UC System",
"types": ["alias"],
"lang" : "en"
},
{
"value" : "University of California System",
"types": ["ror_display", "label"],
"lang" : "en"
},
{
"value" : "Université de Californie",
"types": ["label"],
"lang" : "fr"
}
]
Location information (addresses, country)
The current addresses
field contains data from GeoNames. It contains several sub-fields that contain no values and are therefore not usable, and the ROR team also spends a disproportionate amount of time handling issues with validating and retrieving very granular GeoNames data within addresses
(e.g., geonames_admin2
, nuts_level1
, nuts_level2
, etc.) that users could easily retrieve themselves directly from GeoNames using the GeoNames ID provided in each ROR record. There is also a country
field in each ROR record that duplicates the country information in the addresses
field.
We are therefore removing empty or overly detailed GeoNames sub-fields within the ROR record, removing the country
field, and adding a locations
field that will contain the most important and universally applicable location information. Since most countries do not have states, we are removing the state
and state_code
fields. Users who wish to use location information at the level of the "state" are advised to retrieve it from GeoNames using the geonames_id
.
Identifying organizations without U.S. state information
ROR is a global registry, and most countries do not have an administrative region corresponding to the U.S. state. Note that our analysis shows that there appear to be no records with duplicate name, city, and country in the ROR registry, so users should be able to choose the correct organization based on name, city, and country alone, e.g., "York College, York, United States" (https://ror.org/022jz8688) and "York College, York, United Kingdom" (https://ror.org/04gaeyc40).
Current v1.0 example
"addresses" : [
{
"city" : "Oakland",
"country_geonames_id" : 6252001,
"geonames_city" : {
"city" : "Oakland",
"geonames_admin1" : {
"ascii_name" : "California",
"code" : "US.CA",
"id" : 5332921,
"name" : "California"
},
"geonames_admin2" : {
"ascii_name" : "Alameda County",
"code" : "US.CA.001",
"id" : 5322745,
"name" : "Alameda County"
},
"id" : 5378538,
"license" : {
"attribution" : "Data from geonames.org under a CC-BY 3.0 license",
"license" : "http://creativecommons.org/licenses/by/3.0/"
},
"nuts_level1" : {
"code" : null,
"name" : null
},
"nuts_level2" : {
"code" : null,
"name" : null
},
"nuts_level3" : {
"code" : null,
"name" : null
}
},
"lat" : 37.802168,
"line" : null,
"lng" : -122.271281,
"postcode" : null,
"primary" : false,
"state" : "California",
"state_code" : "US-CA"
}
]
Forthcoming v2.0 example
"locations" : [
{
"geonames_id" : 5378538,
"geonames_details" : {
"country_code" : "US",
"country_name" : "United States",
"lat" : 37.802168,
"lng" : -122.271281,
"name": "Oakland"
}
}
]
Web domain information (links, ip_addresses, email_address, wikipedia_url)
ROR contains four fields related to an organization’s web domain/presence (links
, ip_address
, email_address
and wikipedia_url
), two of which have never had any values (ip_address
and email_address
). URLs in the field links
are inconsistently formatted, and there is no consensus across the data about what value this field should represent (a home page? an About page? something else?).
Further, use cases from the community point to the domain registered to a particular institution (not including protocol, path portions, query parameters, etc.) as being much more useful than a full URL, as it can be unambiguously mapped to records in other services, such as identity management systems. Also, while full URLs may fail to resolve over time, domains are much more persistent, and certainly more persistent than IP addresses, given the rise of public cloud computing.
We are therefore removing the ip_addresses
and email_addresses
fields, adding a domains
field, and creating a links
field to store organization websites and Wikipedia pages.
Current v1.0 example
"email_address" : "",
"ip_addresses" : [],
"links" : [
"http://www.universityofcalifornia.edu/"
],
"wikipedia_url":"http://en.wikipedia.org/wiki/University_of_California"
Forthcoming v2.0 example
"domains" : [
"universityofcalifornia.edu"
],
"links" : [
{
"type": "website",
"value": "http://www.universityofcalifornia.edu/"
},
{
"type": "wikipedia",
"value": "http://en.wikipedia.org/wiki/University_of_California"
}
],
External identifiers (external_ids)
The current structure of the external_ids
field is problematic because it uses the identifier type as a field name. This means that a significant schema change is required to add a new identifier type or remove a deprecated type. ROR currently includes a significant number of records with deprecated external IDs (e.g., CNRS, HESA, and OrgRef) and has received several requests to add new external ID types.
We are therefore revising the external_ids
field to be more flexible, with external identifier schemes expressed as type values rather than field names. We are also removing fields for deprecated identifier schemes.
Current v1.0 example
"external_ids" : {
"FundRef" : {
"all" : [
"100005595",
"100009350",
"100004802",
"100010574",
"100005188",
"100005192"
],
"preferred" : "100005595"
},
"GRID" : {
"all" : "grid.30389.31",
"preferred" : "grid.30389.31"
},
"ISNI" : {
"all" : [
"0000 0001 2348 0690"
],
"preferred" : null
},
"OrgRef" : {
"all" : [
"31921"
],
"preferred" : null
},
"Wikidata" : {
"all" : [
"Q184478"
],
"preferred" : null
}
}
Forthcoming v2.0 example
"external_ids" : [
{
"type" : "fundref",
"all" : [
"100005595",
"100009350",
"100004802",
"100010574",
"100005188",
"100005192"
],
"preferred" : "100005595"
},
{
"type" : "grid",
"all" : [
"grid.30389.31"
],
"preferred" : "grid.30389.31"
},
{
"type" : "isni",
"all" : [
"0000 0001 2348 0690"
],
"preferred" : null
},
Created/last modified dates (new field)
ROR records do not currently contain created or last modified dates, which presents challenges for both data users and ROR staff. For example, data dump users cannot easily ingest only new/updated records: they must either compare a new data dump to a previous data dump in order to identify changes or ingest the entire data set each time a new version is released. Even ROR staff cannot easily identify when a change was introduced to a record based on the record data; staff must consult curation request records and GitHub release history.
We are therefore adding an admin
field to support created and last modified dates as well as other administrative information that may be desirable in the future.
Forthcoming v2.0 example
"admin" : {
"created" : {
"date": "2020-04-25",
"schema_version" : "1.0"
},
"last_modified” : {
"date" : "2022-10-18",
"schema_version" : "2.0"
}
}
Updated 8 months ago