Alternative proposal to handle temporizing from Ward Blonde (MetaRel)

[GoogleCodeExporter]

[Writeup is attachment to this issue]

Hello Alan,

It appears that Barry talked to Vladimir, asking how Metarel could be used 
within the OBO community. He asked to discuss this technically with you. You 
also asked for proposing alternative solutions for the time-indexing problem. 
In attachment I put 16 slides that shows a possible architecture for creating 
semantic precision that can optionally be hidden towards biologist users. The 
presentation was also discussed in an earlier version with Stefan and Mathias.
This proposal might result in a specification for OBO to RDF, in parallel with 
the specification for OBO to OWL. The OWL version would then target 
ontologists, whereas the RDF version can be modularized towards browsing and 
vizualizing applications for biologists.

I hope this might generate some discussion.

Cheers,

Ward

Original issue reported on code.google.com by alanruttenberg@gmail.com on 27 Jul 2012 at 7:37

Attachments:

• BFO_relation_proposal.pdf

[GoogleCodeExporter]

Original comment by alanruttenberg@gmail.com on 27 Jul 2012 at 10:28

• Added labels: Type-BFO2-OWL
• Removed labels: Type-BFO2-Reference

[GoogleCodeExporter]

I also wonder whether dispositions could be expressed in the same way as time 
indexes:

Protein A is located in cellular location B:
- holds at all times
- holds at some times
- is dispositional

The formal meaning could then be expressed in OWL, whereas the RDF model 
remains less complex and more query friendly...

Ward

Original comment by wardblo...@gmail.com on 30 Jul 2012 at 11:03

[GoogleCodeExporter]

Hi Ward,

Leaving aside for the moment the question of whether this is a good idea, I 
don't see how the proposed extension to OBO format can work, as it appears to 
attach 'quantification with time indexing' to classes rather than to 
relationships.

[Term]
id: URI_s
name: biological cell
def: "A biological is a cell that is biological." [WB] is_a: URI_cell ! cell
relationship: URI_p URI_o
has_quantification: at-all-times-all-some

If instead, these tags were used in relation (Typedef) stanzas, then wouldn't 
the proposal be essentially equivalent to the status quo?

Original comment by dosu...@gmail.com on 30 Jul 2012 at 11:20

[GoogleCodeExporter]

Hi,

I admit it is all a bit a cryptic. But I was suggesting a strict binding 
between the relationship tag, followed by the quantification tag. A good syntax 
has to be found for that. Engineering this into the relationship (typedef) 
section of OBO is of course also possible. But that will create the explosion 
of scary-looking relations that this proposal is trying to avoid. The idea 
would be to have these only in OWL. But nothing prohibits to have them, e.g., 
in OBO and OWL, but not in RDF.

Ward

Original comment by wardblo...@gmail.com on 30 Jul 2012 at 11:34

[GoogleCodeExporter]

OK. That makes some sense. Explosion of scary looking relations is clearly a 
potential problem for adoption of BFO2.

For backwards compatibility, there would need to be a default interpretation 
for cases where the quantifier tag is lacking (it seems safe to assume that any 
adoption would be slow and spotty). Have you thought about how this could work?

Original comment by dosu...@gmail.com on 30 Jul 2012 at 12:38

[GoogleCodeExporter]

Hi,

I think this backwards compatibility is a major concern here. Especially if 
there should be an option to hide any enigmatic/scary BFO terminology. At least 
I heard from Barry at ICBO that developers should think on how to hide BFO in 
presenting ontologies to users (biologist users). Only for this reason, I think 
that the good old 'is part of' from BFO1 should continue to be supported also 
in BFO2. At least as a label, if not as a relation. 
The interpretation of such upper-level relations (like 'is part of', 'is 
located in', etc., having no explicit quantification) is a bit a delicate issue 
for me, since I am not into formal ontology. I see several options: 
1) do not give them an interpretation as an ontological relation, but just as a 
modeling artefact, 
2) interprete them as 'for all subjects x, there exists some time t and some 
object y, such that xRy at t'. or 
3) dispositional: 'for all subjects x, there is some subject y, such that x has 
the disposition that xRy holds'

It should be mentioned that not all relations will require quantifier tags. The 
implication of this proposal is that the OBO Format remains very flexible to 
make a lot of errors against BFO, but that a translation towards OWL (DL) will 
reveal the validity. In other words: use OBO for intuitive ontology drafts and 
annotations, OWL for the consistency checking, and RDF for driving applications.

Ward

Original comment by wardblo...@gmail.com on 31 Jul 2012 at 8:17