It should NOT be possible to give sameHgConcept relations to pits with different types

[mmmenno]

By accident, I related an address to a BAG building id. This works, and I think maybe it shouldn't?

[bertspaan]

Correct. Your data is wrong. Core/Graphmalizer does yet check for such mistakes.

[wires]

indeed, shouldn't be possible, but it is more subtle than it seems: you can add a same relation between ids that are not in the graph yet, so we don't know their types yet...

[bertspaan]

We could make system that will disallow those links, such that the respective dataset owners get notified when errors occur.

[mmmenno]

Or a system that simply deletes relations like this - let them figure it out themselves!

Not a big priority, as far as i'm concerned.

[bertspaan]

We currently also allow somebody added the relation Utrecht -> hg:sameHgConcept -> Amsterdam, which would be just as bad for Histograph's integrity.

[mmmenno]

But then, I think maybe it should be possible to say a rm/510640 -> hg:sameHgConcept -> bag/153100000251610, while the first is a hg:Monument and the second a hg:Building.

[sbocconi]

Is this an issue of (not) allowing sameHgConcept relations between PITs of different types? It reminds me of the the owl:sameAs discussion ...
There all is determined by the "semantic strength" of the relation itself (for reasoning), and being sameAs very strong two objects (instances) of different types (classes) cannot be sameAs unless they are also of both types (classes), with all the reasoning consequences of this.

[mmmenno]

I think the best solution might be to change the ontology. Now all types are on the same level, making hg:Monument a subclass of hg:Building would solve this issue.

[tomdemeyer]

Yes. That sounds like a sensible solution. @bert: what are the implications? I figure not too much?

Tom Demeyer
Waag Society
http://waag.org

On 20 Jan 2016, at 15:39, mmmenno notifications@github.com wrote:

I think the best solution might be to change the ontology. Now all types are on the same level, making hg:Monument a subclass of hg:Building would solve this issue.

—
Reply to this email directly or view it on GitHub #74 (comment).

[bertspaan]

Histograph currently has no notion of subclasses. We could add this, though. I know, however, that @wires was strongly opposed to subtypes. Forgot why. @wires, why?

[bertspaan]

And see @wires comment in this issue:

indeed, shouldn't be possible, but it is more subtle than it seems: you can add a same relation between ids that are not in the graph yet, so we don't know their types yet...

[mmmenno]

So, since ...

• you can give a sameAs relation to a non-existing, type-less pit
• sometimes you want a sameAs relation between say a monument and a building
• so far this hasn't been a problem
• you could easily corrupt things by declaring hg:Place Utrecht is sameAs hg:Place Amsterdam as well

... maybe I should just close this issue? Agree?

When this starts to be a problem i gather it won't be difficult to write something to detect and delete sameAs relations between different types (perhaps monument-building excluded)?