That’s why we tend to use RDF, sometimes OWL, for this purpose. In LPG afaik you have to do some hacks like representing literates as nodes, or dumping json as property values to achieve the same. Are there other solutions?

...there are times when we need to sacrifice, o sacrilege, ontological rigour for practical solutions so we can do good in this world.. It needs it.

.. but truth is, we cant really agree fully either. What is your take? I personally believe that discussions like github.com/OBOFoundry/C... (which is certainly what the "What is a phenotype?" discussion would devolve into) lead nowhere; and that yes..

Fringe use case for SSSOM, not for the world :P of course that’s a hugely important use case! Thanks for chatting.

Furthermore, we propose specific modeling solutions for three different categories of entities:

Thank you @piermonn.bsky.social! SSSOM is pretty agnostic to the entity type. When we present about SSSOM (mapping-commons.github.io/sssom/presen...), we usually show the picture below to give a sense of what "things" (symbols) can constitute a mappable entity in the sense of SSSOM..

... , I think yours may be a fringe use case. The point of SSSOM is not primarily to give empower your projects use cases, but to share your mappings that they are re-usable for others, FAIR style. Still, thank you for sharing your cases and giving me an opportunity to learn about it!

Yes, I can see now; While you could express the joining decision in SSSOM like

subject_id: my:table1.col1
predicate_id: skos:exactMatch
object_id: my:table2.col1
mapping_justification: sempav:ManualMappingCuration

(indeed some people do when they are doing data model mappings)...

This is not the fore use case of SSSOM, but if you are up for it we can create an example to see how it would look like. Can you share a gsheet/tsv that contains a column mapping like the one you describe?

… you could use sssom to justify the matching (resolution decisions). I assume your problem also involves transforming data, like concatenation of names etc?

Just to be clear - in your scenario SSSOM only covers the entity resolution part, and only if there is a sane way to identify an individual (I don’t work much with relational database but I guess this would be the primary key). So if two tables in your lake would refer to the same person….