Mapping the relation types from DataCite to Schema.org

[rebeccaringuette]

This issue hosts the development of the mapping from DataCite to Schema.org for relation types. This mapping should tell the user what field in Schema.org to map to based on the relation type.

First, post the correct DataCite xml where all relevant options have been exhausted. Next, post the mapped schema.org result that aligns with this group's format and validates. Also add any modifications needed to the mapping. Then, iterate with those interested towards a final result.

[rebeccaringuette]

The list of relationTypes from DataCite with the proposed mapping of each term to schema.org is below. Some of the relationTypes are not relevant for datasets and are indicated with a "N/A".

• IsCitedBy -> @reverse/citation
• Cites -> citation
• IsSupplementTo -> ???
• IsSupplementedBy -> ???
• IsContinuedBy -> ??? "position" somehow
• Continues -> ??? "position" somehow
• IsDescribedBy -> @reverse/mentions
• Describes -> mentions
• HasMetadata -> subjectOf (see example here from lines 699 to 709)
• IsMetadataFor - N/A
• HasVersion -> prov:hadDerivation, @reverse/isBasedOn
• IsVersionOf -> prov:wasDerivedFrom, isBasedOn
• IsNewVersionOf -> prov:wasDerivedFrom, isBasedOn
• IsPreviousVersionOf -> prov:hadDerivation, @reverse/isBasedOn
• IsPartOf -> isPartOf
• HasPart -> hasPart
• IsPublishedIn -> prov:wasMemberOf
• IsReferencedBy -> @reverse/mentions
• References -> mentions
• IsDocumentedBy -> @reverse/mentions
• Documents -> mentions
• IsCompiledBy - N/A
• Compiles - N/A
• IsVariantFormOf -> prov:wasRevisionOf
• IsOriginalFormOf -> prov:hadRevision
• IsIdenticalTo -> sameas
• IsReviewedBy -> review
• Reviews -> @reverse/review
• IsDerivedFrom -> prov:wasDerivedFrom, isBasedOn
• IsSourceOf -> prov:hadRevision, @reverse/isBasedOn
• IsRequiredBy - N/A?
• Requires - N/A?
• IsObsoletedBy -> prov:wasInvalidatedBy
• Obsoletes -> prov:invalidated
• IsCollectedBy -> prov:wasGeneratedBy/prov:used (see example on lines 528-550)
• Collects - N/A
• IsTranslationOf - N/A
• HasTranslation - N/A

See #270 for how to represent these in DataCite and the proposed mapping structure to Schema.org.

Table 5 of the PROV-O vocabulary was useful for this. Some of these are pure guesses, and some schema or PROV-O terms are used in multiple places. This will render impossible a pure round trip conversion (DataCite -> Schema.org -> DataCite), so please update where a better term is known, otherwise we may be stuck without a good round trip OR we exclude relationTypes from this work where a sufficient schema or PROV-O term cannot be found. Please also comment on relationTypes that you think should be excluded from this list as possibly not relevant to Datasets (e.g., Requires and IsRequiredBy?)

[mbjones]

This looks like a great start. Based on usage in ORE and DataONE, I would request a few differences:

• IsDescribedBy -> subjectOf (synonym of hasMetadata, see SOSO Metadata section)

• Describes -> about -- see SOSO Metadata section

• HasMetadata -> subjectOf (see SOSO guide)

• IsMetadataFor - about (see SOSO guide)

• HasVersion -> prov:wasRevisionOf

• IsVersionOf -> @reverse/prov:wasRevisionOf (I might have these backwards, need to look at the DataCite definitions -- the prov property points at the earlier version)

• IsNewVersionOf -> prov:wasRevisionOf

• IsPreviousVersionOf -> @reverse/isBasedOn

• IsDocumentedBy -> in DataONE we use ore:isDocumentedBy (also in the RDA data packaging recommendation)

• Documents -> in DataONE we use ore:documents

• IsVariantFormOf -> prov:wasRevisionOf (shouldn't these be reversed? plus, is versioning really what is being indicated? I think maybe this refers to stuff like format conversions)

• IsOriginalFormOf -> prov:hadRevision (shouldn't these be reversed? plus, is versioning really what is being indicated? I think maybe this refers to stuff like format conversions)

• IsDerivedFrom -> prov:wasDerivedFrom, isBasedOn

• IsSourceOf -> @reverse/prov:wasDerivedFrom, @reverse/isBasedOn (note that prov:hadRevision is not the inverse property of prov:wasDerivedFrom)

• IsObsoletedBy ->I think invalidatedBy has a different meaning. In dataone, this is a synonym of prov:wasRevisionOf

• Obsoletes -> in DataONE, we would say @reverse/prov:wasRevisionOf

[rebeccaringuette]

Thanks, Matt. Your changes are well-received! A few comments/questions:

• The IsDescribedBy and Describes are often used for things that are not metadata, e.g., https://doi.org/10.5281/zenodo.8253002 which uses that relationship to point to a publication that describes the software and how it was used (e.g., a reference publication). I suspect similar usage exists for data. We should not assume those links are for metadata.
• ORE's last version was in 2008, which I think is before schema.org even existed, so I hesitate on those. How recent was the RDA data packaging recommendation and do you have a link?
• The DataCite definition for IsVersionOf also points to the earlier version. So that seems to make prov:WasRevisionOf the correct choice for IsVersionOf, and vice versa for the HasVersion relation type.
• For IsPreviousVersionOf, you wrote @reverse/isBasedOn, but that is equivalent to prov:wasDerivedFrom, which is used elsewhere. It makes more sense to me to treat this the same as the HasVersion relation type and have it map to @reverse/prov:wasRevisionOf.
• You make a good point on the variant relation types. Any ideas for a better term, or is the revision idea close enough? IsVariantFormOf points to the original version, so prov:wasRevisionOf at least points in the correct direction. Maybe prov:wasDerivedFrom and isBasedOn are closer? Or would similarTo be better?

Generally, I have no objection to the other changes. They make sense to me.
Do you have any ideas on the relation types that have question marks in the first post? Does DataOne have something in place for those?

[mbjones]

The RDA data package rec is here: https://www.rd-alliance.org/groups/research-data-repository-interoperability-wg/outputs/?output=94628

[rebeccaringuette]

Hmmmm. It looks like the BagIt items didn't see much adoption, if any. There are only 2 items there from 7-8 years ago and nothing since. The RDA requirement to include DataCite metadata does further motivate the effort to map between schema.org and DataCite, though.
I see that DataOne generated an example in the document. Does DataOne still use that method, or is the schema.org method preferred? More importantly, is DataOne considering moving away from using terms from ORE due to it not being maintained for almost 20 years?

[mbjones]

Both ORE and BagIt are standards from the library community that are still widely deployed and used, even though the specs themselves have been very stable. DataONE does still use ORE as the basis for our data package manifests (its just an RDF graph as a manifest), and we do support downloading packages in both BagIt0.97 and BagIt 1.0 formats. Other data packaging tools like ROCrate and Dataverse also support it last I checked. In general, I think it is difficult to tell who has adopted RDA recommendations from their site alone, which doesn't meaningfully track usage. In general, I think schema.org is not a very good substitute because the semantics are so ambiguous that is used in very different ways. ORE and json data.package specs have the benefit of much more precise specifications. The RDA spec indicates how a schema.org metadata record (and other metadata formats) can be included in a data package, so they are not mutually exclusive.

[rebeccaringuette]

Thanks.

[rebeccaringuette]

So, the updated list is below with the corrections from above. Did I miss anything? Any objections to just not mapping DataCite's relation types that we cannot find acceptable terms for? Or should we map those to "mentions" or "@reverse/mentions"?

• IsCitedBy -> @reverse/citation
• Cites -> citation
• IsSupplementTo -> ???
• IsSupplementedBy -> ???
• IsContinuedBy -> ??? "position" somehow
• Continues -> ??? "position" somehow
• IsDescribedBy -> @reverse/mentions
• Describes -> mentions
• HasMetadata -> subjectOf (see SOSO guide)
• IsMetadataFor - about (see SOSO guide)
• HasVersion -> @reverse/prov:WasRevisionOf
• IsVersionOf -> prov:WasRevisionOf
• IsNewVersionOf -> prov:WasRevisionOf
• IsPreviousVersionOf -> @reverse/prov:WasRevisionOf
• IsPartOf -> isPartOf
• HasPart -> hasPart
• IsPublishedIn -> prov:wasMemberOf
• IsReferencedBy -> @reverse/mentions
• References -> mentions
• IsDocumentedBy -> ore:isDocumentedBy
• Documents -> ore:documents
• IsCompiledBy - N/A
• Compiles - N/A
• IsVariantFormOf -> ???
• IsOriginalFormOf -> ???
• IsIdenticalTo -> sameas
• IsReviewedBy -> review
• Reviews -> @reverse/review
• IsDerivedFrom -> prov:wasDerivedFrom, isBasedOn
• IsSourceOf -> @reverse/prov:wasDerivedFrom, @reverse/isBasedOn
• IsRequiredBy - N/A?
• Requires - N/A?
• IsObsoletedBy -> @reverse/prov:wasRevisionOf
• Obsoletes -> prov:wasRevisionOf
• IsCollectedBy -> prov:wasGeneratedBy/prov:used (see example on lines 528-550)
• Collects - N/A
• IsTranslationOf - N/A
• HasTranslation - N/A

[rebeccaringuette]

If codemeta gets schema.org to extend the range of the term "documentation" to include "CreativeWorks", then that term and its reverse would be great for the "IsDescribedBy" and its related term in DataCite.