Consider using LinkML

[VladimirAlexiev]

@OR13 @nissimsan
For both Traceability, Edi3, Uncefact you use JSON Schema (plus a bit) as your modeling language.

But that's not enough because it leaves a lot of the semantic aspects loose.

Before trade & logistics had heard of semantics, life science people were doing massive cross-institutional semantic KGs. So check this out:

https://linkml.io/
https://github.com/linkml
Google "linkml" in particular see berkeleybop video etc

Last time I looked (1y ago) it was called BioLinkML and had peculiarities for their case. I think this had now been generalized.

---

Another option could be hackolade.com. It's commercial and doesn't yet have RDF stuff but explains very well the ideas of Polyglot Storage (or Polyglot Persistence)

[nissimsan]

LinkML looks awesome!

[VladimirAlexiev]

• https://linkml.io/linkml/intro/overview.html#we-eat-our-own-dogfood is pretty convincing: it's good enough to describe itself
• Your JSON schemas are not lost, they can be used to bootstrap: https://github.com/linkml/linkml-model-enrichment

[nissimsan]

No volunteers - closing.

[VladimirAlexiev]

@nissimsan @OR13 I agree that switching the tooling in the middle of the river may not be the best thing.

But how do you plan to fix #270?
JSON Schema doesn't have "typed references", so how will you express this most basic semantic thing
"use here this prop, which has that range".