LOBID seems to miss some embedded GND mappings

[nichtich]

See https://lobid.org/gnd/4026894-9 and same GND record in Cocoda: only one embedded mapping is detected but there are more closeMatch, exactMatch and DDC mappings.

[stefandesu]

Good call. Currently, only the sameAs property of the LOBID JSON record is parsed. There are also exactMatch, closeMatch, and I assume more. However, these currently only include the target concept URI, without any information about the target vocabulary (in sameAs, this is included via collection). We could simply include these without toScheme, but most applications would have trouble actually using this, I think.

[nichtich]

The list of target vocabularies is small and each vocabulary has a known URI namespace, these could be hardcoded.

[stefandesu]

The list of target vocabularies is small and each vocabulary has a known URI namespace, these could be hardcoded.

Sounds good. Is there a list of target vocabularies for embedded mappings in GND?

[nichtich]

So far I've seen

• DDC http://dewey.info/class/
• LCSH http://id.loc.gov/authorities/subjects/
• RAMEAU https://data.bnf.fr/ark:/12148/
• BNCF http://purl.org/bncf/tid/
• EMBNE https://datos.bne.es/resource/
• Wikidata http://www.wikidata.org/entity/

This should be enough to start with.

[acka47]

FYI, a full list of enrichments and the linking properties used in GND/EntityFacts can be found at https://wiki.dnb.de/x/TZa5C

[stefandesu]

I'm confused, as most of those listed by @nichtich are not on that list. 🤔 @acka47

[acka47]

Sorry for the confusion. This list at https://wiki.dnb.de/x/TZa5C wasn't the best pointer for this context as it is about the sameAs statements. For the linking sources @nichtich listed other RDF properties are used, see his example https://lobid.org/gnd/4026894-9.

DDC:

"relatedDdcWithDegreeOfDeterminacy2" : [ {
    "id" : "http://dewey.info/class/1--0285/",
    "label" : "http://dewey.info/class/1--0285/"
  } ],
  "relatedDdcWithDegreeOfDeterminacy3" : [ {
    "id" : "http://dewey.info/class/004/",
    "label" : "http://dewey.info/class/004/"
  } ],

LCSH, RAMEAU, BNCF, EMBNE (and STW should also work like this):

"closeMatch" : [ {
    "id" : "http://id.loc.gov/authorities/subjects/sh89003285",
    "label" : "http://id.loc.gov/authorities/subjects/sh89003285"
  }, {
    "id" : "https://data.bnf.fr/ark:/12148/cb11932109b",
    "label" : "https://data.bnf.fr/ark:/12148/cb11932109b"
  }, {
    "id" : "http://purl.org/bncf/tid/1576",
    "label" : "http://purl.org/bncf/tid/1576"
  }, {
    "id" : "https://datos.bne.es/resource/XX525961",
    "label" : "https://datos.bne.es/resource/XX525961"
  } ],

However, at least RAMEAU, are also linked via sameAs.

Searching for SKOS links in lobid, I get back some more sources, e.g. $ curl "https://lobid.org/gnd/search?q=_exists_:exactMatch&size=500" | jq .member[].exactMatch[].label | sort yields concepts in these namespaces:

• http://lod.gesis.org/thesoz
• https://aims.fao.org/aos/agrovoc/ & http://aims.fao.org/aos/agrovoc/
• https://data.bnf.fr/ark:/
• https://id.loc.gov/authorities/names
• https://id.loc.gov/authorities/subjects & http://id.loc.gov/authorities/subjects
• https://id.nlm.nih.gov/mesh/ & https://id.nlm.nih.gov/mesh/
• https://provenienz.gbv.de/T-PRO_Thesaurus_der_Provenienzbegriffe
• https://zbw.eu/stw/descriptor/ & http://zbw.eu/stw/descriptor/

Searching for closeMatch ($ url "https://lobid.org/gnd/search?q=_exists_:closeMatch&size=500" | jq .member[].closeMatch[].label | sort) adds https://purl.org/bncf/tid/, https://purl.org/bncf/tid/, and https://datos.bne.es/resource/.

Interestingly, for some sources (id.loc.gov, agrovoc, mesh, stw, BNCF) both http and https URI schemas can be found which is probaby not intended. (Ping @thoffma.)

[thoffma]

Thank you for pointing this out. We will check the URIs in this regard before the next dump creation in October/November.

[stefandesu]

Implemented and released in v3.4.11. I also updated the Cocoda dev instance, so you can see the results for the example above here: https://coli-conc.gbv.de/cocoda/dev/?fromScheme=http%3A%2F%2Fbartoc.org%2Fen%2Fnode%2F430&from=https%3A%2F%2Fd-nb.info%2Fgnd%2F4026894-9

Possible improvements:

• Check if hardcoded mappings types for different properties are okay: a290b7c#diff-125ad04a29911164a913cbe0bdc8ae76677482a4dcabc3bcd7a58192c50eebf3R103-R110
• http/https URI fixes in data (currently worked around by using uriPattern instead of namespace)
• Question: Is it enough to specify BARTOC URIs for hardcoded vocabularies?
• Find a better solution for the current DDC workaround (the issue is that DDC URIs in GND do not include the DDC version)