Alternative proposal to handle temporizing from Ward Blonde (MetaRel)

[zhengj2007]

From alanruttenberg@gmail.com on July 27, 2012 03:37:16

[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

Attachment: BFO_relation_proposal.pdf

Original issue: http://code.google.com/p/bfo/issues/detail?id=117

[zhengj2007]

From alanruttenberg@gmail.com on July 27, 2012 03:28:20

Owner: alanruttenberg@gmail.com
Labels: -Type-BFO2-Reference Type-BFO2-OWL

[zhengj2007]

From wardblo...@gmail.com on July 30, 2012 04:03:13

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

[zhengj2007]

From dosu...@gmail.com on July 30, 2012 04:20:41

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?

[zhengj2007]

From wardblo...@gmail.com on July 30, 2012 04:34:55

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

[zhengj2007]

From dosu...@gmail.com on July 30, 2012 05:38:31

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?

[zhengj2007]

From wardblo...@gmail.com on July 31, 2012 01:17:51

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