Issue 459: modelling principles - introduction text for crm - version 7.0

posted by Steve on 9/12/2019

This has been sat on my desk for a couple of weeks and finally I have got round to typing up my thoughts.

I think it should be 2 sections: One covering Monotonicity and one covering all the different facets of Extensions. In the attached document I have coloured coded the current paragraphs into these two topics. Blue is extensions and green is monotonicity.

Hope that this helps

Posted by Martin 3/1/2020

Dear All,

I wish you all a Happy New Year, successful and in good health.

Here my attempt to describe the reality concept of the CRM and its relation to a knowledge base. Please comment!!
---------------------------------------------------------------

Reality and Knowledge Bases
The CIDOC CRM is a formal ontology in the sense introduced by N.Guarino [XXXX]. In order to understand the function of a formal ontology for collecting information in research processes about the past that can be shared, connected and integrated into coherent resources, one needs to make the following distinctions:

a) The material reality. For the purpose of the CRM, it is taken as that which is of a substance that can be perceived with senses or instruments, such as people, a forest or a settlement environment, sea, atmosphere, distant celestial or cellular micro structures, including what we assume that could be potentially or theoretically perceived if we could be there, such as the center of Earth or the sun, and all that is past. It is constraint to space and time. What is going on in our minds and produced by our minds is also regarded as part of the material reality, as it becomes materially evident to other people at least by our utterances, behavior and products.
b) The units of description or particulars, i.e., the things and relations as which we distinguish parts of reality when we refer to it, such as Mount Ida, the Taj Mahal,  the formation of China by emperor Qin Shi Huang (秦始皇) in 221BC,Tut-Ankh Amun and his embalmment, Prince Shotoku of Japan sending a mission to China in 607AD, the participation of Socrates in the Battle of Potidaea or the radiocarbon dating of the Iceman Ötzi[1].
A formal ontology, such as the CIDOC CRM, constitutes a controlled language for talking about particulars. I.e., it provides definitions of classes and properties for categorizing particulars as so-called “instances” in a way that their individuation, unity and relevant properties are as unambiguous as possible.  For instance, Tut-Ankh Amun as instance of E21 Person is the real pharaoh from his birth to death, and not extending to his mummy, as follows from the specification of the class E21 Person and its properties in the CRM.

For clarification, the CRM does not take a position against or in favor of the existence of spiritual substance nor of substance not accessible by either senses or instruments, nor does it suggest a materialistic philosophy. However, for practical reasons, it relies on the priority of integrating information based on material evidence available for whatever human experience. The CRM only commits to a unique material reality independent from the observer.

When we provide descriptions of particulars, we need to refer to them by unique names, titles or constructed identifiers, instances of E41 Appellation in the CRM, if the reference should be independent from context, such as reference by pronouns or enumerations of characteristic properties. The appellation itself, and the relation between the appellation and the referred item or relationship, must not be confused with the referred and its identity. Instances of the CRM are the real particulars, not their names. Particulars are approximate individuations, like sections, of parts of reality.

In contrast, a CRM-compatible knowledge base is an information object, instance of E73 Information Object in the CRM. It relates appellations with identifiers of CRM-Concepts in propositions about a described reality. Thereby users, in their capacity of having real-world knowledge and cognition, may be able to relate these propositions to the reality they are meant to characterize, and reason and research about their validity. In other words, the formal instances in a knowledge base are the identifiers, not the real things or phenomena.  Therefore, a knowledge base does not contain knowledge, but represents knowledge of its maintainers, as long as there exist people that can resolve the used identifiers to their referents.

posted by Martin on 26/1/2020

Dear Thanasi,

Thank you very much. I have attended to your comments and one by George in the attached.

I accepted all changes, and put any change from the version 1_23_2020 in yellow, in order to make a new start.

I had thought about adding that Symbolic Objects may reside in the KB, but I think it makes the head of the reader explode. Should be introduced better in the scope note of Symbolic Object. I put the text in violet.

Thanasi, could you write up the texts for diagram 3 and 4? Could be quite similar to the previous texts. If you finish this, I could write up the STV stuff...

Please all to refer to this version _1_26_2020 (or the one from 1-23_2020 if you have made changes in there) from now on.
 

posted by Martin on 26/1/2020

No, I was confusing things. I suggest you, George, put on google docs the section from "We explain these concepts with the help of graphical representations in the next sections"  on until the end, finish your version, and then others take up.

Posted by CEO on 4/2/2020

Dear all,
I am sorry for not participating earlier. I have read the  current final  version of the document.  It may be not always easy to understand even for me. Maybe some parts of the introduction to the "Principles for Modelling Ontologies: A Short Reference Guide could" could be used? 

I will return with a more detailed suggestion tomorow.

Posted by Martin on 6/2/2020

Dear All,

Attached my attempt to bring all bits and comments together in one consolidated version. I have reordered sections according to Pat's and Thanasi's comments.

I have rewritten Carlo's statement about contradictory KBs and reduced to cardinality violations.

I have inserted George's draft about spatial and temporal relations.

I kindly ask Thanasi to check if he would change these two sections.

I have not inserted George's draft about fig 1, but left my own version. I kindly ask George to check, if he would prefer his draft or improve my draft.

I will continue writing the section about spatiotemporal relations.

I kindly ask Nicola to create the promised RDF about Winkelmann, the full story as I refer to it, or just figure 2.

Since we cannot afford during the meeting making minor linguistic optimisations for such a long text, I ask in particular Steve to take over and do that before the meeting.

This text is mission critical for the next official release.

Please let me know ASAP, who can start these final revisions now. Best upload again to Google docs.

Further I attach a graphic designer's inspiration of the Winkelmann time-line (he died in Italy..). Please comment. I'd like to add that to the text, in some further improved version.

Posted Thanasis on by 9/2/2020

Please find attached the text with some comments on George's sections
which apart from some editorial changes I think are good summaries of
how the various properties are meant to be used. I think the text was
cropped in the end. Happy to look at the remaining text if this is the case.

Posted by CEO on 10/2/2020

Dear all,
I have read the document (the latest version). It is in general clear. The part "Introduction to the basic concepts" is ok. We should add a figure 1b with events and actors.

I have some comments to the text in the first part. I have added comments not edited the text as the rule.

Posted by Christian Emil on 10/2/2020

Hi
According to OED obnoxious is defined as "Offensive, objectionable, odious, highly disagreeable. Now esp. (of a person): giving offence, acting objectionably; extremely unpleasant, highly dislikable. (Now the usual sense.)"

The easiest solution is to change this into into a reference t o the more computer science  defintion of Gruber
"An ontology is an explicit specification of a conceptualization. The term is borrowed from philosophy, where an ontology is a systematic account of Existence. For AI systems, what “exists” is that which can be represented. (Gruber 1993, 1)"
and then  write something like ' The CIDOC CRM can also be expressed/formalized as   a theory in FOL, see ....

Posted by Martin on 10/2/2020

Dear Christian-Emil,

There are arguments by Nicola, I would follow, that Gruber's is an underdetermined statement. I definitely like the distinctions Nicola makes in his paper from 1998. I miss the ontological commitment in Gruber's definition. That makes a big difference for me.

I'd rather prefer to be imprecise with "language", rather than talking about "conceptualizations" in an indeterminate way...

I'd rather be more precise about "ontological commitment" to an empirically understood reality, which can falsify, by experiment, a conceptualization.
 

Posted by Martin on 10/2/2020

Dear All,

I will probably never understand the following:

We define the CRM internally in TELOS. Then we create the document, replacing TELOS syntactic sugar with the known format in the document, which is a 1 to 1 mapping, another syntactic sugar. We do this for all releases.

Does this make the CRM semiformal, changing syntactic sugar, or is TELOS semiformal?

Is TELOS not a language? I always thought it is. If it is a language, why is my syntactic equivalent not a language?
Even if FOL has "few semantics", is there a measure how much semantics is needed to become "formal"?

How do you interpret "ontological commitment"???

How can any ontology be exclusively logic without an aboutness about the world?

I understand that the aboutness about the world, is an additional element to the logical form, and not a semiformalism.

Nicola Guarino describes in a naive way the real world to be a sort of LEGO set. That is nice, but he explicitly describes an aboutness about something which is not logic, but exists.

I perceive this world of reference as being non-discrete, but nevertheless endowed with a structure we approximate with the ontology.

Finally, how do you perceive a law of physics, such as Maxwell Equations. Is it formal, is it logical, is it about something that exists, that can be observed and verified experimentally, which must be of a nature that cannot be defined by formulae ("electric")?

Seriously.... 

Posted by Carlo on 11/2/2020

> On 10 Feb 2020, at 19:11, Martin Doerr <martin@ics.forth.gr> wrote:
>
> Dear Carlo,
>
> Thank you for pointing me to that. I usually refer to Nicola's paper about ontologies and information systems 1998.
>
> I am a bit at a loss, why the CRM, once we have formulated it in FOL, and you have declared all the things about constants etc., does not have a language. Could you be more specific?

The paper does in fact provides an ontology in the more technical sense of the word (language plus axioms) although it does not use modal logic. In this sense, it complements the specs. This should be said. Please note that I am not doing this for having one more citation to the paper.
>
> I assumed composing Class and property instances following the CRM axioms constitutes a language. Why not?

It does, but the specs do not say what are the rules for “composing Class and property instances”. These rules are part of the grammar of the language. I assume these rules are very clear in your mind, however, they are not stated anywhere in the specs. In addition, there are the rules for stating the axioms, see below.
>
>
> In my private communications with Nicola he pointed out that he does not care about encoding.

But “encoding” means a lot. It may mean the specific concrete syntax one uses for, e.g. numbers or strings, in which case I agree it’s not so important in an ontology. But if it means the abstract syntax for expressing axioms, then it is of paramount importance. For example, by making the RDF abstract syntax a little bit too liberal, the W3C has created OWL Full, an undecidable monster that can express paradoxical classes such as the class of classes which are not instances of themselves. Of course the CRM specs do not need to revisit the history of logic, they just need to choose a decidable language that is enough expressive for the axioms it contains. See below. 
>
> The question about modal logic is interesting. I think in terms of "possible states of affairs”.

Then you are intrinsically in the realm of possible world semantics, which does not sound surprising since everybody agrees with Carnap that meaning is strongly connected with necessity. You only need to make your thoughts explicit.
>
> So, what should we say?:-)

A quick synthesis of the paper would be my suggestion, it can be condensed in 3 lines. I can provide it if you want, I need some time to harmonise it with the rest of the section, but it has to wait until tomorrow afternoon. 
 

Posted by Christian Emil on 11/2/2020

I think we have a pedagogical issue here. Most peapol think that the TELOS is simply the format used in the definition tekxt of the CRM. We either don't know or  have forgotten that there exist a reasoning mechanism in hte TELOS language/system.
 

Posted by Martin on 11/2/2020

Dear Carlo, Christian-Emil,

Thank you very much for your comments:-). If you, Carlo, would add a few lines about the CRM being a formal ontology in the sense of Guarino's paper from 1998, I would appreciate. I do not remember Nicola's definition to require a specific logical language, syntax or modal logic.

We address with the CRM people that have no idea about these things, and I truely believe that with all the efforts to provide FOL definitions, also the logicians and IT guys can guess the missing details.

Can we be as unprecise as possible, without causing ambiguities of substance?

I would indeed expect that any logical theory established to formulate any ontology would allow for multiple ontologies to conform with the same logical formalism.

For me, of uttermost importance is the interpretation function between the "domain" and the ontology in Nicola's paper, and the fact that he states that the logical theory DOES NOT define the meaning completely, but that it is an approximation to reduce the residual unintended models.

My understanding is, that the scope notes CONSTITUTE the interpretation function. IF the domain is not a naive LEGO set, this interpretation function CANNOT be a logical theory. But this does not make the logical theory "semiformal".

For instance, I have not been convinced so far that describing the axioms of parthood, which you can find in many publications, allows anyone to detect parthood in the real world, and that there exists no other relation with the same axioms, but not being parthood. Parthood implies the axioms, and the axioms select a specific kind of parthood, but parthood cannot be inferred from the axioms, I believe.

Same with "E21 Person", etc...

This goes far beyond Thomas Gruber, and makes a difference to contructivism.

Ontological commitment for me has to do with the interpretation function, but we may look at that in more detail.

You say "Choosing a conceptualisation vs choosing another conceptualisation." What is the criterion? choosing for to do what?

Looking forward to discuss in person....

Posted by Christian Emil on 11/2/2020

To act as the Devil's Advocate: Most readers of the CRM don't care very much about the first order logic  interpretations or interpretations in any other formal deduction systems. That is for the advanced and very skilled readers. 

Most reader try to understand how the CRM can be used to publish open data so that they can be linked easily and efficient.  in the same way people try to use RDF or graph databases as a handy way to make a common and flexible index for heterogeneous databases.

Said so, my view is that the definition of CRM should be formulated in such a way that it can accessible and understandable for  people looking for an relatively easy and exhaustive integration tool, but it should also have a stringent logical definition which make it possible to implement it in an  information system with deduction (also called Knowledge Base).  So we should never make it a necessary requirement  that the readers  have skills in formal logic. If we do so our usergroup will not be very large.  However, the skilled readers should see that the definitions also works on a formal logical level and can easily be implemented in formal systems.

Posted by Christian - Emil on 11/2/2020

Excellent, then what remains is to edit the modeling principles chapter  to make  adequate for the two reader groups. This is pedagogical task and not so easy.

Posted by Christian-Emil on 12/2/2020

I have experienced that most of my colleagues in the humanities get lost at parenthesis level zero when it comes to logical expressions.

It may be easier to write separate texts for the logic geeks and the rest.

Posted by Carlo on 14/2/2020

BTW, I think I now understand Martin's objection on the fact that an ontology needs to have some aboutness about the world. 

My answer is that logical axioms express all the aboutness a computer can understand. For humans, you can use scope notes, i.e., natural language, because humans share the world, have experiences in it, and so can (more or less) understand what you mean by an expression such as (I choose a class at random) "This class comprises all phenomena, such as the instances of E4 Periods, E5 Events and states, which happen over a limited extent in time”. We all know what a phenomenon is, what “happen” means and so on. But computers don’t because they do not have access to the world, so all we can tell them to explain what we mean are the logical axioms.

Posted by Martin on 14/2/2020

On 2/14/2020 7:30 PM, Carlo Meghini at ISTI CNR wrote:
> BTW, I think I now understand Martin's objection on the fact that an ontology needs to have some aboutness about the world.
>
> My answer is that logical axioms express all the aboutness a computer can understand. For humans, you can use scope notes, i.e., natural language, because humans share the world, have experiences in it, and so can (more or less) understand what you mean by an expression such as (I choose a class at random) "This class comprises all phenomena, such as the instances of E4 Periods, E5 Events and states, which happen over a limited extent in time”. We all know what a phenomenon is, what “happen” means and so on. But computers don’t because they do not have access to the world, so all we can tell them to explain what we mean are the logical axioms.
>
> Carlo
>
>
Exactly, but data entry is done by humans:-) for humans....

So, first come people, logic assists...logic needs to conform with experiment. At least in science. If not, change logic until it does.

Isn't it?

Posted by Carlo on 17/2/2020

> Exactly, but data entry is done by humans:-) for humans….

So we need a human-readable/usable version, that’s clear.
>
> So, first come people, logic assists...logic needs to conform with experiment. At least in science. If not, change logic until it does.

I’d say that the logic we have today reflect the work of so many scientists and philosophers that I would think twice before changing it, and after thinking twice, I’d leave it as it is. I speak for myself of course 

I am not sure you got my proposed modification of the first sentence, so I’m re-sending it. I’ve deleted from the attached document all the rest of the chapter to make this email lighter.

Posted by Carlo on 17/2/2020

Thank you Martin, I think we agree. Once someone made a proposal, to “logicize” the scope notes. I think it’s a valid proposal! by doing this we will discover the language the SIG uses to tell the meaning of CRM to the users, and a whole new ontology 

Posted by Martin on 16/3/2020

Please update 7.0 with the attached slide 2,4. Changes: style of multiple properties, P167 added to slide 2, following meeting decision.

Thank you,

Posted by Thanasis on 6/4/2020

I think the footnoted version works. I would just replace "embrace" with "for".

In the virtual meeting of editorial group of CIDOC-CRM version 7.0 on 11/5/2020, the editorial group reviewed Martin's Hw on 30/3/2020 and changed the proposed text to 

"The CIDOC CRM is a formal ontology in the sense introduced by (N. Guarino 1998) [fN 1]. The present document is intended to embrace an audience not specialized in computer science and logic; therefore, it focuses on the informal semantics and on the pragmatics of the CIDOC CRM concepts, offering a detailed discussion of the main traits of the conceptualization underlying the CIDOC CRM through basic usage patterns [fN 2]. The CIDOC CRM aims to assist sharing, connecting and integrating information from research about the past. In order to understand the function of a formal ontology of this kind, one needs to make the following distinctions:

[fN 1]:  Nicola Guarino defines a formal ontology as a specification of a set of named concepts used to describe and approximate a part of reality, plus a first-order logical theory narrowing down the intended meaning of the named concepts.

[fN 2]: For the readers interested in computer science and logic, the syntax and formal semantics employed by the CIDOC CRM are given in (Meghini & Doerr 2018), where the computational aspects are also discussed. "

Editorial Group of CRMbase v.7.0

Posted by Christian Emil on 21/6/2020

Dear all,

Please find a document with the HW for the issues 456 and 459. These issues are about changes in the text of the introduction to the CRM. The document will be used in the virtual ​SIG meeting