how are supertypes assigned?

[structureddynamics]

how are supertypes assigned?

[VladimirAlexiev]

I've loaded umbel.n3 and umbel_reference_concepts.n3 to owlim with owl-horst reasoning. You can query at http://ff-dev.ontotext.com/sparql for a few days:

• 66 are rdf:type umbel:FoodDrink
• 876 are rdfs:subClassOf umbel:FoodDrink, including the above 66

[fgiasson]

@VladimirAlexiev recheck this with the latest umbel 1.50 that we just pushed.

Normally everything should be under FoodDrinks, however this is done by inference. All the sub-class-of relationships are not made explicit in the serialization of the reference structure, but should be inferred from the hierarchy.

[VladimirAlexiev]

Took a look at https://github.com/structureddynamics/UMBEL/blob/master/Typologies/FoodDrink-typology.n3. Overall looks good, with some possible exceptions:

• http://umbel.org/umbel/rc/FishFood rdfs:subClassOf http://umbel.org/umbel/rc/PetFood : surely people eat fish too
• http://umbel.org/umbel/rc/MagnesiumSulfate rdfs:subClassOf http://umbel.org/umbel/rc/Salt-ChemicalSubstance : yes, but it's not an edible salt
• http://umbel.org/umbel/rc/BismuthCitrate rdfs:subClassOf http://umbel.org/umbel/rc/Salt-ChemicalSubstance : it's not an edible salt.
  https://en.wikipedia.org/wiki/Ranitidine_bismuth_citrate is a medication
• http://umbel.org/umbel/rc/Caridina : https://en.wikipedia.org/wiki/Caridina, https://en.wikipedia.org/wiki/Bee_shrimp: no evidence anyone eats them
• http://umbel.org/umbel/rc/PlumTree rdfs:subClassOf http://umbel.org/umbel#FoodDrink : the plum is food, the tree hardly
• http://umbel.org/umbel/rc/Sand-Bottled rdfs:subClassOf http://umbel.org/umbel/rc/BottledProduct : didn't know anyone sells Sand in bottles, but surely it's not food/drink
• http://umbel.org/umbel/rc/Marshmallow rdfs:subClassOf http://umbel.org/umbel/rc/FoodIngredientOnly: kids eat them flamed over open fire. Even if you consider it "FoodIngredientOnly", it's not of the same kind as http://umbel.org/umbel/rc/ArtificialFoodColoring

[mkbergman]

@VladimirAlexiev,

We welcome any errors you find with these assignments. Clearly, any misassignments undercut the usefulness of UMBEL. Though we know the system is unlikely to ever be perfect, we are committed to make it as clean as possible.

Fred is correct that inference is used for the final structure assignments, but we test the structure with bottom-up (direct) assignments as well. Any misassignments you find can be readily tested (and, hopefully, corrected!).

Thanks!