BFO 2020 "Binaryized" Relations

[Anthony Petosa]

(For your reference, see https://www.youtube.com/watch?v=IDs7Pthdows.)

While BFO Continuants are always time-indexed, BFO 2020 introduces "binaryized" relations that hold between Continuant particulars. Consider, for instance, the "continuant_part_of_at_some_time" and "continuant_part_of_at_all_times" binaryized relations. It is not always the case that some particular is a continuant part of some other particular at all times.

For example, the current low 'E' string on Harry's HD-28 Martin acoustic guitar is a continuant part of his guitar only until such time as Harry changes the strings. This is a case for the "continuant_part_of_at_some_time" relation.

BFO 2020 binaryizes ternary relations by creating "_at_some_time" and "_at_all_times" forms (e.g., "continuant_part_of_at _some_time" and "continuant_part_of_at_all_times") for what is otherwise a ternary relation (e.g., "continuant part of at t"). In BFO 2020 OWL, the only difference between these two binaryized relations is that the latter is asserted to be transitive. Otherwise, the domain and range values are "Continuant" with both forms.

How do these binaryized relations enforce a temporal component in OWL given their OWL descriptions do not express a temporal component?

[Alan Ruttenberg]

I'm not exactly sure what is meant by "enforce a temporal component in OWL", so perhaps you could clarify that.

The purpose of the binarized relationships is twofold. First, they have a clear translation to BFO FOL. If they ignore time then they don't have a translation. The OBO binary part-of relation between continuants has no interpretation in BFO.  Second, they provide some temporal reasoning, which is better than none. The specific reasoning they support is transitivity in the at-all-times case, the implication that at-all-times implies at-some-time, and a straightforward inverse relation in the case of at-some-time.

 

They are not perfect. They don't let you say that x is part of something now and part of something later in a way that supports transitivity. The inverses of the at-all-times relations are complicated by the fact that the all-times means the times at which the subject of the relation exists. A has-part-at-all-times defined in a similar way isn't the inverse of part-of-at-all-times. That was the case for the old class-level relations as well. Finally, they don't support the permanent generic relation - a relation that says that at all times subject exists there is a relation to some instance of a class, but not necessarily the same instance at each time.

There may be other ways of bringing time into the OWL, but this is the only way that has been worked through at this point.

Alan

--
You received this message because you are subscribed to the Google Groups "BFO Discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bfo-discuss...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/67fdc76e-3bfc-427c-b622-0c9eefdff523n%40googlegroups.com.

[Anthony Petosa]

As a follow-up, I attached a PDF containing the following Protege screenshots:

• 'continuant part of at some time' description
• 'continuant part of at all times' description
• Ontograf visualization of these relations including the 'Arc Types' panel

The human-readable definitions express a temporal component, but the actual OWL object type property implementations lack a temporal component. More specifically, the '_at_some_time' form lacks a temporal component. The '_at_all_times' is understood to be valid for all temporal values, in which case I assume there is no need to express a temporal relation in OWL.

Suppose we instantiate two 'Continuant' OWL Individuals, 'A' and 'B', and then assert 'A' continuant_part_of_at_some_time 'B'. Since this relation has no temporal component in its OWL object type property description, then how do we enforce this temporal requirement?

The 'exists_at' BFO 2020 relation seems to be a reasonable choice. Suppose we add this additional fact into an BFO 2020-compliant OWL model.

"A exists_at Now", where 'A' is an instance of 'Continuant' and 'Now' is an instance of 'Temporal Region'.

The 'exists_at' relation does not specify domain and range values, which means their values default to owl:Thing. The 'exists_at' elucidation reads as follows:

"(Elucidation) exists at is a relation between a particular and some temporal region at which the particular exists"

Not quite sure why 'particular' is used here as opposed to 'Continuant'. However, the elucidation intends a 'particular' to map to some prescribed time.

If it is true that the BFO 2020 "_at_some_time" relations do not enforce temporality, then why include them? Why not simply assert additional facts in an OWL model that enforce temporality, either at some time or at all times, in the manner suggested by way of the 'A exists_at Now' example?

[Anthony Petosa]

Your post arrived during the time I added to this thread.

>  I'm not exactly sure what is meant by "enforce a temporal component in OWL", so perhaps you could clarify that.

I hope my example offers clarification. Take the 'continuant_part_of_at_some_time' relation, which domain & range values are 'Continuant'. In what way does this OWL implementation allow one to assert that "Mary's left arm" is a '"continuant_part_of_at_some_time" "Mary's body" from Jun 03, 1980 (her birth date) to Nov 18, 2005 (when she unfortunately lost her left arm during a car accident)?

[Anthony Petosa]

The BFO 2020 "bfo-2020-terms.xlsx" Excel workbook "exists at" entry states this relation's domain & range values as "entity" and "temporal region", respectively. The BFO 2020 OWL model does not assert these values on that relation, and I assume this was an oversight. Here are the Excel workbook examples presented for this relation.

"First World War exists at 1914-1916, Mexico exists at January 1, 2000."

Given this statement,

"Mary's left arm" is a '"continuant_part_of_at_some_time" "Mary's body" from Jun 03, 1980 (her birth date) to Nov 18, 2005 (when she unfortunately lost her left arm during a car accident)

, unless there is another OWL-specific implementable solution to include temporality in this binaryized relation, then I do not see how BFO 2020 OWL's "continuant_part_of_at_some_time" relation alone suffices. I think these facts collectively are necessary to assert this statement in OWL. (Note: For the sake of simplicity, I used the BFO universals and not some domain-specific universals.)

"Duration of time that Mary's left arm is a part of Mary's body" instance_of (i.e., rdf:type in OWL) "Temporal Region"

"Mary's left arm" instance_of (i.e., rdf:type in OWL) "Object"

"Mary's body" instance_of (i.e., rdf:type in OWL) "Object Aggregate"

"Mary's left arm" "continuant_part_of_at_some_time" "Mary's body"

"Mary's left arm" exists_at "Duration of time that Mary's left arm is a part of Mary's body"

"Mary's body" exists_at ""Duration of time that Mary's left arm is a part of Mary's body"

The last two facts assert that the arm & body both exist at the same fiat temporal slice of time, which satisfies the temporal portion of the "continuant_part_of_at_some_time" elucidation.