Two Comments on BioTop
Stephen Larson
I have a couple comments on BioTop. I am pleased that there is a
formalized attempt to flesh out the BFO for the purposes of being a
definitive upper level ontology. However, there are a couple issues
that confuse me.
1. The BFO definition of the necessary and sufficient conditions
to qualify as an IndependentContinuant is: snap:FiatObjectPart or
snap:Object or snap:ObjectAggregate or snap:ObjectBoundary or
snap:Site. However, BioTop adds Atom, PolyatomicEntity, and
SubatomicParticle as additional IndependentContinuants that are
siblings to the above classes. When building our ontology under BFO
we assumed that it was incorrect to add additional siblings to
IndependentContinuant, and instead categorized all our
IndependentContinuants under one of the existing sub-categories. This
might be my ignorance of OWL, but wouldn't running an inferencing
engine on instances of Atom, PolyatomicEntity, or SubatomicParticle
actually kick them out of the IndependentContinuant category they do
not satisfy both the necessary and sufficient conditions for being an
IndependentContinuant? And even if not, do you really mean to suggest
that Atoms, PolyatomicEntities and SubatomicParticles are not Objects
or ObjectAggregates, but something fundamentally different? I thought
the point of those BFO sub-categories were that all independent
continuants must be fit into one of them.
2. I notice that the "inheresIn" relation and its inverse are the
only relations that have their domain and range specified. Are there
plans to specify this for the other relations? It seems like
narrowing the domain and the range for relations like componentOf is
pretty crucial to defining what it means. For example, can a
Continuant be a component of an Occurrent? Currently, that is
allowed, but it seems pretty nonsensical to me.
Thanks,
Stephen
Stefan Schulz
>
> 1. The BFO definition of the necessary and sufficient conditions
> to qualify as an IndependentContinuant is: snap:FiatObjectPart or
> snap:Object or snap:ObjectAggregate or snap:ObjectBoundary or
> snap:Site. However, BioTop adds Atom, PolyatomicEntity, and
> SubatomicParticle as additional IndependentContinuants that are
> siblings to the above classes. When building our ontology under BFO
> we assumed that it was incorrect to add additional siblings to
> IndependentContinuant, and instead categorized all our
> IndependentContinuants under one of the existing sub-categories.
I commented on this issued earlier in the BFO list and exchanged some
arguments with Pierre Grenon.
The reason we are currently ignoring the subcategories of
IndependentContinuant is that they seem to be granularity-dependent,
at least the definition of ObjectAggregate:
"An independent continuant that is a mereological sum of separate
objects and possesses non-connected boundaries."
The crucial point here is how " separate objects" and "non-connected
boundaries" should be understood on a molecular level. What is
separated on a fine-grained view may appear connected in a
coarse-grained view.
BioTop is designed neutral to granularity as much as possible, and so
we refrained from using granularity-dependent BFO categories.
In my point of view, a realist toplevel ontology must be
granularity-independent, and so I wonder BFO uses the ill-defined
notion of connectedness and separatedness for partitioning the class
"Dependent Continunant".
> This
> might be my ignorance of OWL, but wouldn't running an inferencing
> engine on instances of Atom, PolyatomicEntity, or SubatomicParticle
> actually kick them out of the IndependentContinuant category they do
> not satisfy both the necessary and sufficient conditions for being an
> IndependentContinuant? And even if not, do you really mean to suggest
> that Atoms, PolyatomicEntities and SubatomicParticles are not Objects
> or ObjectAggregates, but something fundamentally different? I thought
> the point of those BFO sub-categories were that all independent
> continuants must be fit into one of them.
> 2. I notice that the "inheresIn" relation and its inverse are the
> only relations that have their domain and range specified. Are there
> plans to specify this for the other relations? It seems like
> narrowing the domain and the range for relations like componentOf is
> pretty crucial to defining what it means. For example, can a
> Continuant be a component of an Occurrent? Currently, that is
> allowed, but it seems pretty nonsensical to me.
You are right, but most relations are imported from the OBO relation
ontology, so one has to fix it at that level. But we will fix
"component-of".
Best regards,
Stefan