Question about time indexed relations

[toni.c...@dowjones.com]

Hi,

   I'm reading the book "Building Ontologies with Basic Formal Ontology" but when in the chapter about relations there is a talk about time indexed relations like "x continuant-part-of y at t" and this is called a "primitive". But to the best of my knowledge that cannot be encoded in OWL without using a N-Ary class and by greping in the Github repo I've only found entries for that symbol in some lisp files.

   So my question is, how is it defined in the OWL files for BFO? Or how do you use those time indexed concepts in a practical example?

Thanks!

[Pierre Grenon]

Hi, 

I'm not too familiar with the particular write up you are referring to but I understand the book would not assume an OWL encoding as definitive of what BFO counts as primitive. This doesn't mean there may not be at least a partial axiomatisation in OWL. However, such an axiomatisation does not necessarily carry ontological weight ('commitment'). 

Regarding the OWL encoding of BFO, I believe BFO defers to RO. I understand that RO in OWL uses non-temporalised language. In practice, the best way of using this may be to assert facts in time indexed contexts, in which the temporal parameter for the relation is implicitly that of the context. 

On a purely technical level, still assuming that by 'using in a practical context' you mean encoding ground facts with temporalised relations, the simplest strategy is to reify a particular relationship in a rather standard move:

foo (continuant) part of bar at t1 

becomes 

foobart1 part_whole foo ; 

                 part_part bar ; 

                 part_time t1. 

It is purely technical and there is no BFO interpretation for foobart1. The temporalised parthood is a formal relation. The 'thematic role' relations used in reifications are not part of BFO as such either although clearly you need such structural vocabulary if you wish to express facts like this one in OWL.

Perhaps this has been worked out more standardly, in which case Ill defer to a more up-to-date answer. Others will correct me.

Cheers

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.
To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/0a58f7dc-58ba-4b50-8835-00ee755d1b78%40googlegroups.com.

[Toni Cebrián]

I think your answer partly solved my question. So in general, I don't understand how "Chapter 7 Ontology of Relations" fits when developing my own ontology. To give an example, in page 135:

We can now identify the following primitive instance-level relations and their

definitions:

• c instance_of C at t. This is a primitive relation obtaining at a specific time between

a continuant instance c and a continuant universal C when the former instantiates the

latter at that time. For example: Fido instance_of Labrador Retriever at the present. 

so say, that I have instances for Fido, a class for Labrador Retriever extending continuant, and present an instance using Time OWL. How would you state fact "Fido instance_of Labrador Retriever at the present" using BFO? 

In any of the ontologies stated in https://basic-formal-ontology.org/users.html can I see an example of above definition?

Bear in mind that I'm just learning ontology stuff and that for instance your sentence "such an axiomatisation does not necessarily carry ontological weight ('commitment')" didn't make any sense to me, so maybe the problem is that I'm not able to grasp the context on which Chapter 7 of that book happens.

Thanks for your answer, BTW.

Toni

On Saturday, August 17, 2019 at 1:13:22 AM UTC+2, pierre...@gmail.com wrote:

Hi, 

I'm not too familiar with the particular write up you are referring to but I understand the book would not assume an OWL encoding as definitive of what BFO counts as primitive. This doesn't mean there may not be at least a partial axiomatisation in OWL. However, such an axiomatisation does not necessarily carry ontological weight ('commitment'). 

Regarding the OWL encoding of BFO, I believe BFO defers to RO. I understand that RO in OWL uses non-temporalised language. In practice, the best way of using this may be to assert facts in time indexed contexts, in which the temporal parameter for the relation is implicitly that of the context. 

On a purely technical level, still assuming that by 'using in a practical context' you mean encoding ground facts with temporalised relations, the simplest strategy is to reify a particular relationship in a rather standard move:

foo (continuant) part of bar at t1 

becomes 

foobart1 part_whole foo ; 

                 part_part bar ; 

                 part_time t1. 

It is purely technical and there is no BFO interpretation for foobart1. The temporalised parthood is a formal relation. The 'thematic role' relations used in reifications are not part of BFO as such either although clearly you need such structural vocabulary if you wish to express facts like this one in OWL.

Perhaps this has been worked out more standardly, in which case Ill defer to a more up-to-date answer. Others will correct me.

Cheers

Pierre

Le ven. 16 août 2019 à 18:57, <toni....@dowjones.com> a écrit :

Hi,

   I'm reading the book "Building Ontologies with Basic Formal Ontology" but when in the chapter about relations there is a talk about time indexed relations like "x continuant-part-of y at t" and this is called a "primitive". But to the best of my knowledge that cannot be encoded in OWL without using a N-Ary class and by greping in the Github repo I've only found entries for that symbol in some lisp files.

   So my question is, how is it defined in the OWL files for BFO? Or how do you use those time indexed concepts in a practical example?

Thanks!

--
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.

[Pierre Grenon]

Hi Toni,

glad it helped a bit, sorry you still are looking for an answer.

It would help if you gave some background on where you are coming
from, why you are learning ontology and what it means to you. This is
not a test but simply, ontology has a long history, is
multidisciplinary and has multiple aspects from which people may
approach it and it is very hard to pinpoint the right level of details
or the relevant things to say.

Meanwhile, I can only attempt to answer the question of how to
possibly best use the book (bear in mind I'm only talking as a reader
and potential user of the book). First, again, the book is not giving
a standard OWL implementation of BFO. (I am not sure whether this
makes sense to you or whether you assume that an ontology is something
written in OWL.) What this means is that you do not use the book by
trying to take a term and add it to an OWL ontology. If you want to
use OWL, you have to think and make choices of how to encode certain
things and this may create trade-offs, for example, some things you
won't encode or some things you will need to add to your
representation.

The book is attempting to expose BFO --- it tells you 'this is BFO (to
some particular extend and to some particular level of details)'. BFO
in turn tells you 'these are the sorts of things that exist'. The
book does its job using primarily natural language and a first order
logic style of syntax for making it a bit more precise. It does not
expose the full articulated meaning of every concept but this is not
so relevant. The point to note is that you cannot use such input to
derive a uniquely prescribed encoding in OWL. What the book gives you
instead is an underlying intended semantics and, to be charitable, it
allows any encoding that fits the bill and preserves this semantics.

So, you are right in your observation that the book does not tell you
how to use OWL in the case of temporalised relations but in general it
does not tell you how to use OWL in any other case either.

I personally do not know of any encoding of temporalised relations in
the ontologies listed as using BFO but I have not scourged these
files, so this is no answer to the question of how people do it. I
would be very interested to find out indeed. I would look for
ontologies that seem to deal with the way material objects change
(say, anatomical growth) or maybe attributes that are very time
dependent (e.g., clinical measurement) --- you will likely not find
any examples in ontologies that aim at setting vocabulary for
relatively static categories (e.g., census stuff) and those that are
very event or process oriented will take advantage of occurrent
language.

As you indicate in your original post, you could use an n-ary
formalisation for temporalised relations. My argument is that this is
merely an apparent complication due to a language choice (OWL). You
could do precisely that with Fido. (There is a slightly technical
issue in that there is no language in BFO for talking of the present
as such, but you could introduce it as part of language but this has
to do with the peculiar choice of words in the example.)

Typically, in the context of the book, BFO is used to build
hierarchies (if you look at these ontologies you are pointing to, it
is mostly what they are). In practice, the most valuable things for
this are what the book calls 'universal-universal' relations and the
others are in the background, they secure some underlying semantics.
The way the book would be used here by somebody building a hierarchy
in OWL, say, would be to say:
OK, I'm going to use subclass as BFO wants me to understand is_a (in
other words, I am going to assert rdfs:subClassOf when I mean is_a).
This means, in particular, that I'm going to understand instantiation
a la BFO (following the temporal variant for continuants)--- I am not
going to use it in assertions, however. That is a very weak thing to
say and do. I am ruling out cases in which all As were Bs yesterday
but today some As aren't Bs. But I am not going to express this in
OWL. What I'll express is "A rdfs:subClassOf B" or whatever I chose to
encode is_a. This is weak and there is wishful thinking involved in
believing that I have actually encoded what I meant.
In my experience, all the ontologies you cited do that. It is fine.
They mean to say 'that's what I mean'.

If now I want to use a temporalised relation in an OWL encoding. I
have varied choices. What the book asks me to do is to be consistent
with the underlying semantics. So if I use the predicates from my
first response, I bind myself to say 'oh wait part_whole rdfs:range
bfo:Continunat, btw'. As I understand, that's what the book asks me to
do if I chose this route.

In cases like these, when there is no prescription, I would argue that
anything goes. I just can't go back and say 'aha BFO is missing a
part_whole relation' or 'where does this foobart1 fall in BFO?
shouldn't there be a class of ContinuantParthoodRelationshipAtATime
instantiated by things like foobart1?'. The answer is no. I have
introduced formal vocabulary in a specific linguistic context
(RDF/OWL) and it is to encode a given theory, but the element of my
language do not add to that theory. This is what I meant with the
remark on ontological commitment. Ontological commitment is a
technical concept that can be made very complicated, I used the phrase
to mean 'whatever BFO tells you exists'. The book aims at clarifying
whatever BFO tells you exists, the way you formulate this or encode it
does not add to what BFO tells.

This later bit is potentially controversial and you can see it as
pedantic depending on where you are coming from, it does depend on
broader philosophical stances taken on the nature of ontology,
ontologies, what they do, how they are built and what they are used
for in some cases --- I think the book expresses some of these and
some others differently than I would. However, if somebody takes a
stance in which OWL is a language that is to be naively interpreted in
a way that each class term is a category and each predicate term is a
relation and that is what the essence of your ontology is, then it is
very likely that what I am saying does not make a lot of sense. It's
OK if that's your case, I hope the explanations help clarifying some
of the uneasy feeling then.

Does this help more with understanding how Chapter 7 helps?

On the specifics, do you have ideas on how you would encode "Fido is a
Labrador today" in OWL according to BFO?

Best,
Pierre

> 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/8865d094-49cf-4417-b328-eff3af4d3bcd%40googlegroups.com.

[Barry Smith]

Regarding the BFO treatment of temporalized relations

The new ISO version of BFO, which will be publicly available very soon, provides an approach to temporalized relations based on work by Alan Ruttenberg. The relevant formal content is available here. For an introduction see the tutorial videos here, especially part 1 (from 18:45) and part 6. See also the corresponding slides here, 

A more extensive version of this material will also be presented at the Tutorial on Top-Level Ontologies (ISO/IEC 21833) to be held at the JOWO meeting in Graz on September 24. Video from this will be released shortly thereafter. 

To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/CAAkf34wNDwy5uPfad%2Bf%2BD%3DTksLtrWytCFk4-JOR60NAtFkOUJA%40mail.gmail.com.

[petosa...@gmail.com]

Greetings, all.

 

I look forward to reading the new BFO ISO version, particularly as it applies to temporalized relations. The “RO” ontology, as far as I can remember, always was considered an interim solution for BFO 2 temporalized relations. I am a bit wary of the decision to continue using the “RO” ontology as part of BFO 2. In fairness, I have not viewed the latest incarnation of the “RO” ontology, but I’ve seen plenty of references to past versions indicating it is not a “best fit” for BFO.

 

With respect to OWL, given RDF is built upon a two-tuple atomic data structure (i.e., the ‘subject-predicate-object’ construct, which itself is a pattern imposed upon a mathematically underlying directed graph) and whereas BFO universals follow a realist ontological perspective, whereby its Continuants are ‘time-indexed’ and its Occurrents are inherently time-dependent, the RDF(/OWL) two-tuple construct fails to model reality directly. That is “time” is a key factor in reality. Ideally, a machine-consumable atomic data structure would innately address time, since our reality does not exist without it. Therefore, real temporal constraints, as in “John is now 27 years old”, have to account for temporal artifacts in roundabout ways. For instance, one can place John’s age behind an OWL B-node implemented as a Datatype Property Restriction. However, this temporal artifact is not publicly exposed, because a B-node’s identifier does not resolve to an IRI; its identifier is generated by an OWL modeling tool, and, therefore, is not directly accessible over the World Wide Web (see https://www.w3.org/TR/rdf11-concepts/#section-blank-nodes). This seems contradictory to the original Open World Assumption of the “web”, because such a node must be explicitly asserted as fact.

 

Then, of course, we must consider the trans-ontological temporal relationships that exist between universals on the BFO Continuant side (the 3D Perspectivalist view of reality) with the BFO Occurrent side (the 4D Perspectivalist view of reality). For example, given the aforementioned RDF/OWL data structure constraint, how can one best model an Object – Process specific dependence relation, both at the “universal” level and at the “particular” level? For example, to say that a Process particular specifically depends upon one or more Object particulars, requires either a temporal instant or a temporal range within which the Object particular(s) participate in the process, and the temporal component, if a range, can be either contiguous or discontiguous.

 

> For an introduction see the tutorial videos here…

Unfortunately, the audio is very poorly recorded. I’d love to give the video a view once the audio issue is rectified.

 

 

Regards,

 

Anthony Petosa

To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/CAN_82STu7MnMPw4AjH97XU7MTtuOsD%3DEA0N0hb4zF1XwpUV7zA%40mail.gmail.com.

[Alan Ruttenberg]

One strategy to have at least some temporal expressivity is to define part-of-at-some-time and part-of-at-all-times for continuants. The all-times version is transitive, the some-times not. The at-all-times version is a subproperty of the at-some-time version. This is what I did for the ISO version. occurrent-part-of is binary so no problem with OWL for that.

As for instance-of, for the moment the only thing we have is to define rdf-type as an at-all-times relation thought we continue to experiment with different strategies for representation in OWL.

An important limitation of this approach is that the Class-class pattern as described in Relations in Biomedical Ontologies doesn't work for cases where entities change class over time. For instance A part of B with A, B considered classes translates into: anything that is an instance of A at a time is part of an instance of B at that time. 

Another strategy, also limited, is to try to express some of the material entity temporal stuff in terms of history. For instance you might say that the history of the life of A is (occurrent) part of the history of the life of B to mean that A exists for a shorter period of time than B and during that time A exists it is part of A.

I'll try to find some time to see if the audio of the tutorial can be improved.

Regards,

Alan

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/8865d094-49cf-4417-b328-eff3af4d3bcd%40googlegroups.com.

[Zander Cage]

testing