Uncertain Relations

40 views
Skip to first unread message

R Ilin

unread,
Sep 3, 2019, 2:18:08 PM9/3/19
to BFO Discuss
Hello,

I am new to BFO and was wondering if I could get some recommendations.  Does BFO allow the use of uncertain relations? For example, I see an animal and I am not sure if that is a duck or a goose.  Could I say something like "This animal may-be a duck"?  Is there a conventional way of handling this situation?

Thanks,

Roman.

Barry Smith

unread,
Sep 3, 2019, 2:46:54 PM9/3/19
to bfo-d...@googlegroups.com
This is not the final story, but you may find some food for thought here

--
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/1c8cce83-5ffd-4ac1-ab34-50e82341277e%40googlegroups.com.

R Ilin

unread,
Sep 3, 2019, 3:36:32 PM9/3/19
to BFO Discuss
Thank you for the reference.  I found this chapter online  https://philpapers.org/archive/NEUMPA.pdf ,  is this the same text or would you recommend getting the book from the library? 


On Tuesday, September 3, 2019 at 2:46:54 PM UTC-4, Barry Smith wrote:
This is not the final story, but you may find some food for thought here

On Tue, Sep 3, 2019 at 2:18 PM R Ilin <rili...@gmail.com> wrote:
Hello,

I am new to BFO and was wondering if I could get some recommendations.  Does BFO allow the use of uncertain relations? For example, I see an animal and I am not sure if that is a duck or a goose.  Could I say something like "This animal may-be a duck"?  Is there a conventional way of handling this situation?

Thanks,

Roman.

--
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-d...@googlegroups.com.

Alan Ruttenberg

unread,
Sep 3, 2019, 5:44:17 PM9/3/19
to bfo-d...@googlegroups.com
Why not say: this animal is a (duck or goose)? This can be said in both OWL and FOL.
Alan

On Tue, Sep 3, 2019 at 2:18 PM R Ilin <rili...@gmail.com> wrote:
--
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.

R Ilin

unread,
Sep 4, 2019, 8:58:15 AM9/4/19
to BFO Discuss
Alan, Barry, let me reformulate my example.  I think the way I wrote it was not adequate.  The situation is actually an output of a classifier.  For example I have a camera taking images of the scene, and after an object is detected, the classifier returns the list of possible classes with probabilities attached.  So, my example really is to say something like this:

The detected object is-an-instance-of ( duck with probability 80% OR goose with probability 20%)

If the classifier has 100 possible classes I'll have a lot of terms. If I am only interested in ducks, I need a statement 

The detected object is-an-instance-of duck with probability 80%, 

which could be stated as 

The detected object may-be-an-instance-of a duck 

So the question is:  is it OK to use such "may-be" relation? Plus, I am not sure how I could attach the probability to my relation - is there a way to do this in OWL?  

Is there a better/standard way to handle this situation? 

Thanks,

Roman.

Pierre Grenon

unread,
Sep 4, 2019, 9:48:47 AM9/4/19
to BFO Discuss
Hi Roman, 

Welcome.

In the context of BFO, if you see something and think with our without certainty that this something is a duck, that very fact has nothing to do whether that thing is a duck but only to do with what your thought is like. In that sense, there is no degree of (un)certainty that is attached to a relation in BFO. 

The standard way of approaching modal and epistemic qualification (but see Barry's reference on canonicity for a possibly different take) is to use BFO as factually oriented --- you use it to state facts--- and have an additional mechanism in your language allowing for epistemic or modal qualification. 

Compare:
- mickey is a duck
- it is likely that Mickey is a duck
- the likelihood of mickey being a duck is .7

BFO only intends to cater for the first part. 

I'll follow up on your rephrased question.

Cheers,
Pierre


--

Pierre Grenon

unread,
Sep 4, 2019, 10:11:39 AM9/4/19
to BFO Discuss
There is nothing like this in BFO. But it is not strictly speaking an ontological issue as much as a language issue. 

You could introduce modally qualified binary predicate if you wished. That would not really help you with what you intend to record however. It is cleaner to use a factual vocabularg and represenf the epistemic weights on your statements. 

If ypu want to do this with RDF, what you may want to do is have a class of classifications and your own ontology. (I assume you may want to allow for different models producing different results.)

Then you record:

myKB:classif123 rdf:type myOnt:Classification.
classif123 myOnt:classifiedObject myKB:ObjectA.
myKB:classif123 myOnt:classifiedAs domain:Duck.
myKB:classif123 myOnt:classifiedBy myKB:Model12345.
myKB:classif123 myOnt:score ".7".

and whatever else is relevant.

This is totally independent of BFO. 

You can hook up to BFO to have your domain ontology integrated or abstracted over. So from the above you infer whatever you want from domain:Duck. 

You could want to do something more discerning. Instead of classifying an object, you could classify a scene. In that sense you may be able to classifg potential relatikns, for example, a head is part of a bird. You could express

there is 80% likelihood that object a is part of object b 

by having a different class of classifications in which you would classify the relation. But applying the naive stub above you could have:

myOnt:Classification.
classif123 myOnt:classifiedObject myKB:RelationBetweenAAndB.
myKB:classif123 myOnt:classifiedAs bfo:isPartOf.

In that way, you can use BFO as domain vocabulary used to express statements --- that you only describe but do not assert. In that approach, BFO relatikns remain either absolutely true or absolutely false and you ascribe a probability to facts. To my mind it is much cleaner and also closer to what is actually being intended.

Best regards,
Pierre





--
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.

Hunter, Larry

unread,
Sep 4, 2019, 10:16:08 AM9/4/19
to bfo-d...@googlegroups.com

This sounds to me as it should be modeled similarly to the way that OBI handles the output of an instrument in a scientific experiment. The output of the classifier process is a class and a number which is a probability. See http://obi-ontology.org/

Ludger Jansen

unread,
Sep 4, 2019, 10:19:47 AM9/4/19
to bfo-d...@googlegroups.com

Dear Roman and all

Interesting discussion so far!

It sounds as if the "real" entity in question is not the bird, but the detection process, that has certain participants (a computer and a bird) and a certain output ... IAO and OBI could be of help to integrate this into a BFO framework (the Information Artifact Ontology and the Ontology for Biomedical Investigations, that is).

Best
Ludger

To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/CAAkf34xxzbaj2emKncV8S8jjDbOLWQmfODiHEfcjpD-dYQpUig%40mail.gmail.com.
-- 
PD Dr. Ludger Jansen
Institut für Philosophie
Universität Rostock
D-18051 Rostock

petosa...@gmail.com

unread,
Sep 4, 2019, 12:51:52 PM9/4/19
to bfo-d...@googlegroups.com

I suppose this depends upon the purpose of Basic Formal Ontology. From https://basic-formal-ontology.org/:

 

“The Basic Formal Ontology (BFO) is a small, upper level ontology that is designed for use in supporting information retrieval, analysis and integration in scientific and other domains.”

 

This statement does not capture the ontological realism undertones of BFO. From https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3104413/ (see Section 1 – “The methodology of ontological realism”, Subsection 1.1 – “The goal of ontology development”)

 

“Ontologies are created to serve multiple goals, including support for more effective retrieval of data and for different sorts of reasoning. Here we focus on ontologies created to foster consistency in the ways scientific results are described for purposes of more effective integration of scientific data – ontologies, therefore, that serve strategies to counteract the many tendencies leading to ad hoc and non-interoperable coding of data, and thus to the formation of data silos… The realist methodology is based on the idea that the most effective way to ensure mutual consistency of ontologies over time and to ensure that ontologies are maintained in such a way as to keep pace with advances in empirical research is to view ontologies as representations of the reality that is described by science. This is the fundamental principle of ontological realism.”

 

The second excerpt paints a more thorough picture of BFO’s purpose. If one views BFO’s goal only in terms  of the first excerpt, then I believe this is a disservice to Basic Formal Ontology, reducing its OWL and FOL representations and those of its derivative OWL and FOL representations to a hodge-podge of defined classes.

 

I believe that allowing defined classes in a realist ontology representation detracts from the empirically-driven foundation of precise & concise ontological representations of “things” that exist in realty. One can easily create set theoretic structures in OWL and FOL that are entirely accurate from a FOL viewpoint but are otherwise absurd in that they have no bearing on reality. For this reason, I question “ontology” definitions such as the following:

 

“ontology = def. a representational artifact, comprising a taxonomy as proper part whose representations are intended to designate some combination of universals, defined classes, and certain relations between them” (see http://marte.aslab.upm.es/redmine/files/dmsf/p_oasys/170119123026_238_Arp_-_Building_Ontologies_with_Basic_Formal_Ontology.pdf, Section 1 – “What Is an Ontology?”, Subsection – “Introduction”, Paragraph 2).

 

This definition comingles universals with defined classes.

 

In summary, as stated at the top of this post, if you accept BFO in this light, then, yes, the defined class “(duck or goose)”, itself subsumed by “animal”, which in turn presumably is subsumed by BFO’s “Object”, is a perfectly logical construct. However, one can just as easily assert the following logical construct as a subtype of BFO’s “Object” universal:

 

(duck or goose or doorknob or hammer)

 

By not imposing (a) a rigid empirical and scientific view of reality within the current realm of human knowledge and (b) a monohierchical view of taxonomy, we can create a plethora of information silos fueled by arbitrary defined classes.

 

With a realist view of ontology, a “duck” universal (a.k.a., type) is clearly defined by science, as are “goose”, “doorknob” and “hammer” universals. The inability to determine what something is at some point in time, as may be the case when one sees an object at a distance in the desert at twelve noon through the distorted haze of an atmospheric disturbance, is irrelevant to ontological realism and should not inform ontology development. “Duck”, “goose”, “doorknob” and “hammer” universals are mutually exclusive. There is no such “thing” in reality, for instance, as a “(duck or goose)”.

R Ilin

unread,
Sep 6, 2019, 8:08:21 AM9/6/19
to BFO Discuss
Thank you everyone for interesting feedback.  I think that the suggestion by Pierre makes sense, which is to create a mid-level ontology for Classification rooted in BFO. I am planning to do this.  I am also wondering if the Common Core Ontologies already have something like this.

Roman.

On Tuesday, September 3, 2019 at 2:18:08 PM UTC-4, R Ilin wrote:

Steve Black

unread,
Apr 7, 2021, 4:39:50 PMApr 7
to BFO Discuss
Hi Roman - Did you have any luck in creating the mid-level ontology for Classification rooted in BFO? I'm working on a classification-type project and would greatly appreciate any ideas, resources, and/or inspiration you could provide. Thanks!

R Ilin

unread,
Apr 7, 2021, 4:50:50 PMApr 7
to bfo-d...@googlegroups.com
No, unfortunately I have not created it. Got distracted by something and did not come back to it yet. I had some initial ideas based on the email chain in this discussion and I can share them with you but I did not take it far so if you read the discussion carefully maybe it can help you get started.  

--
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.


--
Roman ILIN

Steve Black

unread,
Apr 19, 2021, 5:39:12 AMApr 19
to BFO Discuss
I read through the discussion and it was definitely helpful. And I'd appreciate any initial ideas you are open to sharing. The more context, insight and inspiration I can gather at this stage, the better! Feel free to share directly with me at steve...@gmail.com, if you prefer. Thanks!!
Reply all
Reply to author
Forward
0 new messages