Re: [topbraid-users] Punning and SHACL

[Irene Polikoff]

People model for different reasons and there can be different types of models.

While I am quite convinced that Bold Eagle as a specie is different from a set 
of all occurrences of the individual bold eagles and, thus, each deserves their 
own URI, I can't prevent you from designing your models the way you chose. If 
you feel that there are cases where your needs are best met by punning, it is a 
design choice that I assume you would make. 

The original question was about SHACL and whether it would have some inherent 
issues with this approach. The answer was no.

If I am asked, my recommendation would be to avoid punning. This recommendation 
is not specific to TQ tooling, it is a general modeling recommendation. Having 
had now nearly 20 years to think about and work on such modeling issues 
independently and with colleagues whose judgement I respect, I believe punning 
mixes concerns and causes more problems than it solves. 

However, as I said, everyone is free to make their choices. Besides, there can 
always be exceptions to any rule.

> On Sep 8, 2018, at 11:48 PM, Rob Atkinson  wrote:
> 
> 
> Without contradicting anything Irene says (which is obviously the gold truth 
> for the TQ environment) I'd also like to offer an additional architectural 
> perspective (cos thats my day job) that may also be useful, and i'm always 
> seeking to identify best practices here...
> 
> A few separate concerns here:
> 
> 1. We need to be careful with statements like "mean different things" - I 
> think this is a classic case of the httpRange14 issue - i.e. distinguishing 
> between identifiers of "things" vs identifiers of "representations of things" 
>  Representations have shapes, i.e. sets of properties - and this is a design 
> choice about how to model a thing
> 2. Under the Open World Assumption, anyone can say anything about any topic 
> using whatever representation they choose - so this means its another design 
> decision (at a high level of your architecture) how to manage closure of 
> graphs in a distributed environment.  (TQ is largely concerned with your 
> local node - and supports a particular set of closure patterns, but this is 
> all about collapsing the open world into a tractable sub graph _that you 
> choose_ 
> 3. The example given is actually a strong example of why (IMHO) some systems 
> may need to support OWL punning ...
> 
> The model for the concept of "Endangered Species" is inconsistent with poor 
> old "Harry" - which usefully tells us the model is broken or not applied 
> properly to Harry ...
> 
> in fact there is no such thing as an Endangered Species as a class of all 
> individuals.  "Endangered" is a status applied explicitly to a species within 
> a certain spatio-temporal context (usually a date of a politically negotiated 
> determination within a given jurisdiction),  Scientifically derived concepts 
> of endangered may also co-exist - and different evidence and modelling may 
> suggest that a designation is more or less appropriate - a species may be 
> endangered by any rational scientific assessment, bit not declared so, or 
> declared so and  actually extinct. There is usually a "extinct in the wild" 
> explicit status too - so the set of eagles is actually the "set of eagles in 
> the wild" - which poor Harry isnt.  There is nothing wrong with the model of 
> Bald Eagle - it can be instance of species and a class of a set of 
> individuals. Its just not a subclass of Endangered Species according to the 
> model implied in the example. Endagered Species is a different thing with a 
> relationship to a Species, and individuals may or may not be instances of 
> both.
> 
> Of course, you may manage your context  to only include a given jurisdiction 
> (or scientific viewpoints) view - in which case the model is correct but the 
> assignment of Harry to a specimen assumed to be "in the wild" is incorrect. 
> So you probably need to handle instances and classes in many cases - 
> 
> (Having done some modelling of biodiversity and Web scale citizen science 
> interoperability concerns, this is easy for me to spot - and IMHO 
> demonstrates the potential power of the semantic approach)
> 
> 
> 
> 
> On Saturday, 8 September 2018 22:32:12 UTC+3, Irene Polikoff wrote:
> Punning is the use of the same word to mean different things. 
> 
> In modeling, it is traditionally used to refer to the approach where the same 
> resource is used both as an instance of some class (other than the pre-built 
> RDFS/OWL/SHACL classes) as well as declared to be a subclass of some class. I 
> am excluding the pre-built classes from this because that takes us into a 
> topic of meta modeling which can be seen as related, but is somewhat 
> different.
> 
> My e-mail was about this topic and the differences between a concept and a 
> set of occurrences of this concept in the world. 
> 
> When a modeler defines a class they make statements that must be true for all 
> class members and they 

Re: [topbraid-users] Punning and SHACL

[Rob Atkinson]

There is another obvious Use Case...

as skos:Concept is used to classify an instance of a more general class.
Over time it becomes necessary to model the sub set of instances of this
general class that share this classification - so the concept hasnt
changed, but under the OWA and AAA principles we can model it with further
axioms. Often such distinctions are necessary across lines of governance -
one domain only cares about a high level classification whereas another
needs to model in more detail the behaviour of subclasses.

I feel Irene's comments relate more to the (natural) requirement to have
alternative frame based models (shapes - i.e. choices of sets of
properties) for the same thing.  In the W3C Data Exchange Working Group we
are looking to formalise the concept of "profiles" and
content-negotiation-by-profile, so there is a mechanism to keep different
views separate (and conveniently discoverable and accessible)  without
forcing rather arbitrary (and difficult to describe and navigate) subtle
variations in defining the underlying concept to try to keep them distinct.






On Sat, 8 Sep 2018 at 17:55 dprice  wrote:

> To be clear, “punning” is not a modelling concept. It’s just the
> implementation approach that DL reasoners take to handle a class being a
> member of another class. They pretend that there are two separate things (a
> class and an individual) with the same URI - thus the “pun".
>
> I’ve seen many models where classes with members that are also members of
> other classes. If it makes sense in your domain of interest, don’t fear
> using it.
>
> A simple example is “Pump”, the noun, as used in an oil platform. “Pump"
> has physical objects that are its members. However, “Pump" is also a member
> of the class "Rotating Mechanical Equipment Class”, which is a subclass of
> “Mechanical Equipment Class”, and which has other members like “Fan” and
> “Centrifuge”. “Pump” is also a member of the class "NORSOK
> Z-CR-002 Equipment Class” which are classes approved in a specific oil and
> gas standard.
>
> Another similar example is an ontology about the types of things of
> interest to a company that wants to specify, for example, which attributes
> are required to be recorded when a regular survey is done of a specific
> type of asset.  A class called TypeOfRoadAsset might have members Junction,
> Bridge, Motorway, etc. which would then have members “Junction 2 on the GB
> M1”, “London Bridge” and “GB M1” as members. A class called
> TypeOfRoadAttribute might have members Pavement Depth, Date of Build, etc.
> and then a class called AttributeOfRoadAsset would link the assets and
> attributes with members like "Motorway Pavement Depth” that are specified
> as being “mandatory” attributes to be recorded in the yearly road survey.
>
> Both of these examples are kinds of concepts found in real ontologies that
> are in operational use today. This kind of approach is common in the
> manufacturing and engineering asset management industries.
>
> Cheers,
> David
>
>
> On 8 Sep 2018, at 15:03, Adam Kimball  wrote:
>
> Thanks for the help so far.
>
> My intuition suggests the punning is not the right solution, too.  Your
> email, Irene, helps solidify that.  There is no doubt that the properties
> associated with a skos:Concept instance would be radically different from
> an owl:Class instance as you demonstrate with Flu and Audi Q6.
>
> -Adam
>
> On Friday, September 7, 2018 at 6:43:15 PM UTC-6, Irene Polikoff wrote:
>>
>> Adam,
>>
>> Yes, there should be no issue with SHACL.
>>
>> Having said this, I am yet to come across a situation where punning is
>> the best or even the right approach.
>>
>> Typically, when one puns, they really mean to describe two different
>> things that, in a natural language, are referred to using same term.
>> Natural language is informal because when people use the same name to talk
>> about two conceptually different things, they understand from the context
>> what is being talkied about.
>>
>> For example, let’s consider cars and car models. I could decide Audi Q6
>> to be an instance of a class Car Model (or its subclass Audi Car Model). If
>> so, this resource would have certain properties. Or I could decide it to be
>> a class. In the later case, my intention would be for the instances of the
>> class to be the actual cars that have VIN numbers, date of manufacturing,
>> owners that drive them, etc. None of these properties are appropriate for
>> the car model. Depending on what I am trying to express, while the resource
>> name/label could be the same, the meaning would be quite different. Since
>> these resources have different meaning and different properties, they
>> should be separate.
>>
>> Similarly, one could talk about a flu as an instance of a disease and a
>> flu as a set of all occurrences this disease - thus, a class or a set of
>> things, but a set of conceptually different things. Flu as an instance of a
>> disease has typically symptoms, commonly 

Re: [topbraid-users] Punning and SHACL

[dprice]

To be clear, “punning” is not a modelling concept. It’s just the implementation 
approach that DL reasoners take to handle a class being a member of another 
class. They pretend that there are two separate things (a class and an 
individual) with the same URI - thus the “pun".

I’ve seen many models where classes with members that are also members of other 
classes. If it makes sense in your domain of interest, don’t fear using it.

A simple example is “Pump”, the noun, as used in an oil platform. “Pump" has 
physical objects that are its members. However, “Pump" is also a member of the 
class "Rotating Mechanical Equipment Class”, which is a subclass of “Mechanical 
Equipment Class”, and which has other members like “Fan” and “Centrifuge”. 
“Pump” is also a member of the class "NORSOK Z-CR-002 Equipment Class” which 
are classes approved in a specific oil and gas standard.

Another similar example is an ontology about the types of things of interest to 
a company that wants to specify, for example, which attributes are required to 
be recorded when a regular survey is done of a specific type of asset.  A class 
called TypeOfRoadAsset might have members Junction, Bridge, Motorway, etc. 
which would then have members “Junction 2 on the GB M1”, “London Bridge” and 
“GB M1” as members. A class called TypeOfRoadAttribute might have members 
Pavement Depth, Date of Build, etc. and then a class called 
AttributeOfRoadAsset would link the assets and attributes with members like 
"Motorway Pavement Depth” that are specified as being “mandatory” attributes to 
be recorded in the yearly road survey.

Both of these examples are kinds of concepts found in real ontologies that are 
in operational use today. This kind of approach is common in the manufacturing 
and engineering asset management industries.

Cheers,
David

> On 8 Sep 2018, at 15:03, Adam Kimball  wrote:
> 
> Thanks for the help so far.
> 
> My intuition suggests the punning is not the right solution, too.  Your 
> email, Irene, helps solidify that.  There is no doubt that the properties 
> associated with a skos:Concept instance would be radically different from an 
> owl:Class instance as you demonstrate with Flu and Audi Q6.  
> 
> -Adam
> 
> On Friday, September 7, 2018 at 6:43:15 PM UTC-6, Irene Polikoff wrote:
> Adam,
> 
> Yes, there should be no issue with SHACL.
> 
> Having said this, I am yet to come across a situation where punning is the 
> best or even the right approach. 
> 
> Typically, when one puns, they really mean to describe two different things 
> that, in a natural language, are referred to using same term. Natural 
> language is informal because when people use the same name to talk about two 
> conceptually different things, they understand from the context what is being 
> talkied about.
> 
> For example, let’s consider cars and car models. I could decide Audi Q6 to be 
> an instance of a class Car Model (or its subclass Audi Car Model). If so, 
> this resource would have certain properties. Or I could decide it to be a 
> class. In the later case, my intention would be for the instances of the 
> class to be the actual cars that have VIN numbers, date of manufacturing, 
> owners that drive them, etc. None of these properties are appropriate for the 
> car model. Depending on what I am trying to express, while the resource 
> name/label could be the same, the meaning would be quite different. Since 
> these resources have different meaning and different properties, they should 
> be separate.
> 
> Similarly, one could talk about a flu as an instance of a disease and a flu 
> as a set of all occurrences this disease - thus, a class or a set of things, 
> but a set of conceptually different things. Flu as an instance of a disease 
> has typically symptoms, commonly prescribed treatments, may be a person who 
> first discovered it, etc. A flu that someone had let’s say a year ago is a 
> member of the disease occurrences class. It has a start date, an end date, 
> patient that had the disease, etc. These properties are appropriate for the 
> members of the class disease occurrence, but not for the class disease. It 
> was probably treated, but not necessarily using typically prescribed 
> treatments. May be some of them, but definitely not all. There is a 
> relationship between an instance of a disease and a class of decease 
> occurrences (and between a car model and a class of cars), but they are not 
> the same things.
> 
> Barry Smith has hundreds of slides explaining the difference - not that I 
> think hundreds of slides are necessary to communicate this. 
> 
> Of course, I do not know what your situation and needs are. Just thought I’ll 
> share my perspective on this common modeling question.
> 
> Regards,
> 
> Irene.
> 
>> On Sep 7, 2018, at 6:16 PM, Adam Kimball > > wrote:
>> 
>> I've run into a few different scenarios where I would like something to be 
>> both a class and an instance.  For me, this happens when I model 

Re: [topbraid-users] Punning and SHACL

[Adam Kimball]

Thanks for the help so far.

My intuition suggests the punning is not the right solution, too.  Your 
email, Irene, helps solidify that.  There is no doubt that the properties 
associated with a skos:Concept instance would be radically different from 
an owl:Class instance as you demonstrate with Flu and Audi Q6.  

-Adam

On Friday, September 7, 2018 at 6:43:15 PM UTC-6, Irene Polikoff wrote:
>
> Adam,
>
> Yes, there should be no issue with SHACL.
>
> Having said this, I am yet to come across a situation where punning is the 
> best or even the right approach. 
>
> Typically, when one puns, they really mean to describe two different 
> things that, in a natural language, are referred to using same term. 
> Natural language is informal because when people use the same name to talk 
> about two conceptually different things, they understand from the context 
> what is being talkied about.
>
> For example, let’s consider cars and car models. I could decide Audi Q6 to 
> be an instance of a class Car Model (or its subclass Audi Car Model). If 
> so, this resource would have certain properties. Or I could decide it to be 
> a class. In the later case, my intention would be for the instances of the 
> class to be the actual cars that have VIN numbers, date of manufacturing, 
> owners that drive them, etc. None of these properties are appropriate for 
> the car model. Depending on what I am trying to express, while the resource 
> name/label could be the same, the meaning would be quite different. Since 
> these resources have different meaning and different properties, they 
> should be separate.
>
> Similarly, one could talk about a flu as an instance of a disease and a 
> flu as a set of all occurrences this disease - thus, a class or a set of 
> things, but a set of conceptually different things. Flu as an instance of a 
> disease has typically symptoms, commonly prescribed treatments, may be a 
> person who first discovered it, etc. A flu that someone had let’s say a 
> year ago is a member of the disease occurrences class. It has a start date, 
> an end date, patient that had the disease, etc. These properties are 
> appropriate for the members of the class disease occurrence, but not for 
> the class disease. It was probably treated, but not necessarily using 
> typically prescribed treatments. May be some of them, but definitely not 
> all. There is a relationship between an instance of a disease and a class 
> of decease occurrences (and between a car model and a class of cars), but 
> they are not the same things.
>
> Barry Smith has hundreds of slides explaining the difference - not that I 
> think hundreds of slides are necessary to communicate this. 
>
> Of course, I do not know what your situation and needs are. Just thought 
> I’ll share my perspective on this common modeling question.
>
> Regards,
>
> Irene.
>
> On Sep 7, 2018, at 6:16 PM, Adam Kimball  > wrote:
>
> I've run into a few different scenarios where I would like something to be 
> both a class and an instance.  For me, this happens when I model something 
> as a skos:Concept because I want to refer to the abstract Thing but later 
> have a need to refer to individual Things.  In the past, I have always been 
> freaked out by this problem so I've found ways around it that aren't very 
> elegant.  But I'm working on a new project now where I might Stop Worrying 
> About Owl, and Learn to Love SHACL.  
>
> I can't imagine why or how the concept of punning could cause problems in 
> SHACL but I think this list would have the people who could provide a 
> definitive answer.
>
> So, can I pun freely?
>
> -Adam
>
> -- 
> You received this message because you are subscribed to the Google Groups 
> "TopBraid Suite Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to topbraid-user...@googlegroups.com .
> For more options, visit https://groups.google.com/d/optout.
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to topbraid-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [topbraid-users] Punning and SHACL

[Irene Polikoff]

Adam,

Yes, there should be no issue with SHACL.

Having said this, I am yet to come across a situation where punning is the best 
or even the right approach. 

Typically, when one puns, they really mean to describe two different things 
that, in a natural language, are referred to using same term. Natural language 
is informal because when people use the same name to talk about two 
conceptually different things, they understand from the context what is being 
talkied about.

For example, let’s consider cars and car models. I could decide Audi Q6 to be 
an instance of a class Car Model (or its subclass Audi Car Model). If so, this 
resource would have certain properties. Or I could decide it to be a class. In 
the later case, my intention would be for the instances of the class to be the 
actual cars that have VIN numbers, date of manufacturing, owners that drive 
them, etc. None of these properties are appropriate for the car model. 
Depending on what I am trying to express, while the resource name/label could 
be the same, the meaning would be quite different. Since these resources have 
different meaning and different properties, they should be separate.

Similarly, one could talk about a flu as an instance of a disease and a flu as 
a set of all occurrences this disease - thus, a class or a set of things, but a 
set of conceptually different things. Flu as an instance of a disease has 
typically symptoms, commonly prescribed treatments, may be a person who first 
discovered it, etc. A flu that someone had let’s say a year ago is a member of 
the disease occurrences class. It has a start date, an end date, patient that 
had the disease, etc. These properties are appropriate for the members of the 
class disease occurrence, but not for the class disease. It was probably 
treated, but not necessarily using typically prescribed treatments. May be some 
of them, but definitely not all. There is a relationship between an instance of 
a disease and a class of decease occurrences (and between a car model and a 
class of cars), but they are not the same things.

Barry Smith has hundreds of slides explaining the difference - not that I think 
hundreds of slides are necessary to communicate this. 

Of course, I do not know what your situation and needs are. Just thought I’ll 
share my perspective on this common modeling question.

Regards,

Irene.

> On Sep 7, 2018, at 6:16 PM, Adam Kimball  wrote:
> 
> I've run into a few different scenarios where I would like something to be 
> both a class and an instance.  For me, this happens when I model something as 
> a skos:Concept because I want to refer to the abstract Thing but later have a 
> need to refer to individual Things.  In the past, I have always been freaked 
> out by this problem so I've found ways around it that aren't very elegant.  
> But I'm working on a new project now where I might Stop Worrying About Owl, 
> and Learn to Love SHACL.  
> 
> I can't imagine why or how the concept of punning could cause problems in 
> SHACL but I think this list would have the people who could provide a 
> definitive answer.
> 
> So, can I pun freely?
> 
> -Adam
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "TopBraid Suite Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to topbraid-users+unsubscr...@googlegroups.com 
> .
> For more options, visit https://groups.google.com/d/optout 
> .

-- 
You received this message because you are subscribed to the Google Groups 
"TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to topbraid-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [topbraid-users] Punning and SHACL

[Holger Knublauch]

On 8/09/2018 8:16 AM, Adam Kimball wrote:

I've run into a few different scenarios where I would like something 
to be both a class and an instance.  For me, this happens when I model 
something as a skos:Concept because I want to refer to the abstract 
Thing but later have a need to refer to individual Things.  In the 
past, I have always been freaked out by this problem so I've found 
ways around it that aren't very elegant.  But I'm working on a new 
project now where I might Stop Worrying About Owl, and Learn to Love 
SHACL.


I can't imagine why or how the concept of punning could cause problems 
in SHACL but I think this list would have the people who could provide 
a definitive answer.


So, can I pun freely?


Yes. SHACL is based on RDF graphs, no other hidden assumptions.

Holger




-Adam
--
You received this message because you are subscribed to the Google 
Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send 
an email to topbraid-users+unsubscr...@googlegroups.com 
.

For more options, visit https://groups.google.com/d/optout.


--
You received this message because you are subscribed to the Google Groups "TopBraid 
Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to topbraid-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.