search

FAIRsFAIR / FAIRSemantics

MIT License
7 stars 1 forks source link

P-Rec. 13: Crosswalks, mappings and bridging between semantic artefacts should be documented, published and curated. (D2.5) #40

Open GCoen1 opened 3 years ago

GCoen1 commented 3 years ago

Mappings, crosswalks, content negotiations, and semantic bridges discussed in the previous recommendations (P-Rec. 12) should be made publicly available to allow the reuse by others. Mappings are semantic artefacts in themselves, and therefore should be shared and published in semantic repositories, following the recommendations for such artefacts (for example in respect of GUPRI and metadata).

Sharing these resources in a standardised way will improve interoperability. The main requirement is for such mappings to be machine readable for reuse purposes (P-Rec 11) , and to be described with minimum metadata for human interaction.

matentzn commented 3 years ago

Not sure whether mappings should be published as RDF or some such.

This is how an SSSOM mapping file looks like: https://github.com/EnvironmentOntology/environmental-exposure-ontology/blob/master/mappings/ecto.sssom.tsv

which can be mapped into OWL, but they are probably more useful to the wider community in table form then if they are somewhere in RDF where they have to be queried.. I guess what I am saying is that "published in semantic repositories" does not seem to be necessary for mappings in all cases..

dr-shorthair commented 2 years ago

Could you clarify if this recommendation includes all cross-references, or just mappings? Some semantic models will include references to elements from other 'vocabularies' with specific, narrow semantics, e.g the NVS P01 'semantic model' refers to ChEBI for chemical substances. It would be useful if FAIR semantics also has an explicit recommendation to use existing well-established and well-governed vocabularies in cross-references, in preference to creating new collections or vocabularies.

matentzn commented 2 years ago

While we have some recommended predicates in SSSOM, the spec is entirely open to include any kind of mapping relation, including mereological (part of) and other kinds, however in practice, we do recommend using only standard relationships. The only idiosyncratic ones you will find in our domain is the oio:hasDbXref, which is a bit of a historical accident, but we are recommending now to move over to SKOS mapping relations for most purposes.