Issue 429: P72 has Language

Posted by Franco Niccolucci on 26/8/2019

Dear George, all

I think that there are two issues (at least) here.

The first one concerns the identity criteria of this class. This discussion started from an issue related to the latter. In this case the grouping of English speakers, for example, is identified as “those people whose bio states so”. It does not matter if they really speak/spoke English of not, this concerns the veridicality of their bio, which is another story.

So the grouping of English speakers is precisely identified. This is not always the case.

This issue is a particular case of a more general issue concerning fiat vs bona-fide objects, to use the terminology introduced by Smith and Varzi about geographical (but not only) objects. As you may remember, fiat ones have precise boundaries, bona-fide don’t. For groupings, belongingness has the same alternatives, and in most cases what we may call “fiat belongingness” is based on a formal definition, like a listing, mathematical criteria, a decree and so on. There are thus groupings for which it is easy (feasible?) to assess belongingness, others for which it is not, others for which it is unclear. The crm-sig mailing list is an example of a fiat group defined by listing, as is the group of the citizens of Italy at the time I am writing this email, defined by the law and recorded in the civil registry. 
Nationality - mentioned in the E74 scope note - could belong the uncertain case: if you consider nationality as the formal status of being citizen of a country, it is a fiat criterion. But there may be cases in which the nationality may be uncertain. I don’t want to make examples of today as they may be politically sensitive, but if you had asked in 1861 to people from Venice their nationality they would answer “Italian” although their formal nationality was "Austro-Hungarian”. Thanks to the principle of self-determination, the number of such cases is much rarer today than it was in the 19th century, with a few notable exceptions that we all have in mind. However, 99.999% of the cases refer to formal nationality so the above is just a pedantic discussion.

Language(s) spoken is much more difficult to assess: what turns the bona-fide boundary between speakers and non-speakers into a fiat one in this case? A certificate issued by a school? Self-assessment? I think that the case that raised this discussion may be easily solved as I mentioned above. But I would be cautious to use it in other cases.

For the second issue, modelling this grouping as an E74, I understand George’s concern about the use of E74 Group, which is a subclass of E39 Actor and thus is required to “[collectively] have the potential to perform intentional actions of kinds for which someone may be held responsible”, what seems doubtful for speakers of a language. In my opinion this requirement for intentional actions could be considered in a very broad sense; for language, avoiding sexist terminology in English could be an example - stretching it a bit, I admit. But otherwise, how can we model collectivities like this one and others such as “archaeologists”, “Buddhists” “Real Madrid fans” etc ?

Finally, George’s proposal is nice but addresses only the language issue and not other groupings/features of the same type, i.e. collectivities based on some common characteristic, but not required to be able to collectively perform intentional actions, for example illiterate people.

Posted by Franco Niccolucci on 27/8/2019

Steve,

something for your breakfast tomorrow morning. 

“Knowing” a language is not the same as “using” it. The case started from documentation stating that somebody knows a language, but not reporting any use, which is just potential but not necessarily actual. For example, I know Latin pretty well, but I have very few - if any - opportunities of using it; on the contrary, I do not know Japanese but sometimes say “sayonara” and “arigato” appropriately. In these Portuguese archives I would be correctly recorded as “Latin speaker” but not as “Japanese speaker”.
Your solution instead refers to “using" the language as implied by P16 and would state exactly the opposite.

I share with you the hate for negative searches, for the reasons you clearly explain.

Bene valeas placideque quiescas, Stephane (*)

Francus

(*) in order to enable you in using P16 for my knowledge of Latin

Prof. Franco Niccolucci
Director, VAST-LAB
PIN - U. of Florence
Scientific Coordinator
ARIADNEplus - PARTHENOS

Editor-in-Chief
ACM Journal of Computing and Cultural Heritage (JOCCH) 

Piazza Ciardi 25
59100 Prato, Italy

> Il giorno 26 ago 2019, alle ore 23:32, Stephen Stead <steads@paveprime.org> ha scritto:
>
> Just thinking about this after an interesting game of Railroad Revolution.
> It strike me that it might be useful to consider language as a Conceptual Object and an Actors use of it would be an instance of E7 Activity P2 has type E55 Type {Communication} P16 used specific object E28 Conceptual Object.
> E55 Type {Communication} could be sub-divided into written, spoken, reading etc as necessary.
> The other stuff that Rob mentions is rather different and at first glance looks a lot like the floruit from FRBR which became F51 Pursuit.
> I am concerned about building optimisations of properties that are intended for making searches about negative things like “not known to speak Latin” as this is a nasty place to be: absence of Knowledge versus knowledge of absence……
>  
> Use of a technique is that also the use of an immaterial object?
>  
> Anyway off to bed now. Very interesting question
> TTFN
> SdS
>  

Posted by George Bruseker on 27/8/2019

Dear all,

Sticking to the question of documenting when we have information that someone knew a language or had a skill in a technique, I reiterate that I believe really need a new property and not to use p2 has type. 

p2 has type is a good solution for classifying a kind of phenomena or for specializing a class when it does not require a new relation in the ontology. It's a very useful tool but it does not work for what we need to document here.

The semantics of saying that someone had knowledge in a language can indeed be interpreted as E21 Person p2 has type E55 "English Speaker". It could not, however, be typed E21 person p2 has type E57 Language "English". Why? Because the E57 Type classifies the phenomenon of language not of people. The E55 is relative to the phenomenon it classifies/specializes. People are not language nor vice versa. One of the things we would want to make possible in linking an E21 person to an E57 Language is to create consistent and potentially serendipitous relations between an instance of person and an instance of language. (As one of Rob's examples: the work used E57 English and the person who encountered it E5 was a knower of E57 English, ergo, they could but did not necessarily read it!) This would not be facilitated by saying E21 Person p2 has type E55 "English Speaker" because there are no given semantic connections between the instance "English Speaker" which classified a person as a kind and the instance E57 Language "English" which classifies linguistic phenomena. 

The semantic intent, I would argue, in the schemas that document fields like language and technique is often not to say that this person is of type "English Speaker" or even of type "Painter" but that they have/had knowledge of English (linguistic phenomenon) or Painting (technical phenomenon). Because someone knows or uses a technique does make them someone who would generally be classified (with regards to official documentation) as being an exemplar of that language/technique. So as Rob is not necessarily a 'French Speaker" though he knows French, George is not necessarily a Painter, though he may have a knowledge of painting notable enough to document. (It is counterfactual, I don't even have this knowledge but for lack of a better example)

Human beings are an objective phenomenon that can be witnessed and have certain behaviours and potentials which other phenomena do not, one of which includes the ability to know. We should be able to document this objective phenomenon because it falls within scope. The kind of knowledge in question is not an act of knowing (temporal) but the result of having learned and now acquired a new understanding which allows the human being to act in the world in a new skillful way in certain situations. This knowledge remains, more or less present, in the knower without any particular activation once they have acquired it (forgetting and rustiness not withstanding). It is simply one of their properties. 

It's all a long winded way of saying that we need a relation between E21 Person (at least) to indicate that they have a knowledge. There should be a binary property for this (which could then be extended) which allows one to make the simple statement, A knows B. This would not be a sub property of P2 has type, but a new property. I'm not sure if it would have an existing superproperty. My original suggestion would be to stick to language and then go for a super class, although the question of technique also arises. 

The other issues Rob and Franco raise about documenting fiat groups/sets are very important but perhaps we could make them another discussion and issue (when it comes time to formulating something particular for voting on at SIG). 

About the idea of making language a conceptual object, I think we would have to have a lot of discussion and reflection on that, because it seems like a large metaphysical issue. Language is obviously very particular to human being, Aristotle called us the rational (logos) animal. But it is not clear that logos is the invention of human being or that it can be said to be something that we can use in a utilitarian way like a pot or a mould. It seems more like a medium through which certain types of communicative act can/do occur. Anyhow, also a fun discussion but I think having a E21 person "has knowledge or / was use of language" X property could be a modest first step that is semantically robust to a real use case and can be extended by further modelling without likely breaks to monotonic development.

Posted by Thanasis Velios on 27/8/2019

Dear all,

The fact that our documentation systems document a direct relationship between language/technique and person does not mean that a direct relationship is needed in the CRM (we have many examples of direct relationships in documentation systems which do not exist in the CRM, e.g. "the date" of an object). The act of speaking using a specific language or painting using a specific technique can be modelled through the respective activities (and I would argue that agency is not limited by the person's age, i.e. some babies are certainly determined to do things). Modelling through activity allows searching based on the language of a text in relation to the language spoken by a person during events of a specific time-span. I think the queries that Rob mentioned can be answered that way.

I appreciate the difference between the activity of speaking using a language and the mental capacity of a person to speak a language or to hold any knowledge. I am not sure there are any research questions that cannot be answered by modelling this through using activity. To model knowledge is to model deduction for which we already have necessary classes. 

Posted by George Bruseker on 28/8/2019

Hi Christian-Emil et al.,

Regarding language in particular, my argument would be to make a new direct binary relation something like E21 Person pxx 'knew language’ E56 Language. 

This relation, to my mind, would be parallel to E18 p45 consists of E57 Material

There is indeed an event which we normally don’t know anything about (nor have a research interest in) of learning a language, which leads to the instance of E21 Person having a constitutional change in knowledge (Aristotle called it Hexis) whereby they then know a language. I believe this change in knowledge state is not something that changes the being of the individual as such (primary quality) which is what p2 has type would indicate but only creates a modification in the secondary qualities of the person. 

To loosely parallel existing CIDOC CRM modelling, a production event creates an object. In creating it, materials are used and it creates a new instance of Human Made Object. This instance of Human Made Object now consists of an E57 Material like ceramic. So qua what it is made of we say p45 consists of, qua what it functionally is, we say that it p2 has type ‘jug’ for example. p45 is not a sub property of has type because the relation is not one of “being" the material but rather having the substance of material x.

Regarding time problems, the instance of E21 Person did not always know the language. That being said when we declare a relation like ‘knew language’ we state that it was the case that there was a moment of the existence of this E21 Person where the person had the knowledge (had the hexis) of knowing x. It is actually true for the whole lifetime of the entity that at sometime it knew language x just in case in real life at sometime in its life it knew language x. 

I think that in the interest of not endlessly filling up CRMBase, it might be better to put such an addition into CRMSoc. The above suggestion does not mean to argue that we couldn’t or shouldn’t also model learning events or use events with regards to language but rather that there is a basic function that is ontologically correct to assert that a Person knows a language which fits a real world use case. 

Posted by George Bruseker on 28/8/2019

Dear Franco et Al.,

Actually I have no argument against skill, it would be a similar pattern. My worry would be about making either a language or a skill a conceptual object. I see the reason behind the proposal but I'm not sure especially about language being conceived of as a human made object in the crm sense. There I think we more refer to an intentionally created intellectual object with discrete boundaries. I am not convinced this is an appropriate apprehension of language. We come to be in language and reproduce it. A great genius may change it. Mostly it just happens to us and we do not employ it as an object nor are we intentionally aware of it. No one made it. It forms a sort of horizon for communicative axtion.  By using type we avoid the controversy. Skill I would associate with texne in the sense of craft. Craft could be conceived as closer to something like a crm conceptual object but again a craft seems to go beyond any one person or group qua invention so i would find it more comfortable to think of as a type.

So if skill were also to be modelled maybe another binary property

E21person had skill exx skill 
Exx skill isa e55

And then perhaps some super property of both

E21 had knowledge of e55

I would find the ability to express both of these in crm an extremely useful addition instead of creating ad hoc solutions per project.

As to the base issue, I don't mean anything fancy, just whatever is in the basic standard and not a family model. 

Will check out the demonstration vid when back to WiFi!

Posted by Martin on 23/9/2019

Dear All,

I support Christian-Emils proposal. May I remind you, that the CRM has become very stable by a careful bottom-up process. Introducing super-concepts in a rash to cover the whole world is desastrous. If any, we model first language skills from documentation, and then add other skills one by one. Please also compare to the FRBRoo model about extended activities. We describe a potter by doing pottery, not by having the potential. The skill is a potential. CRM is evidence oriented. People speaking a language do it. How do we assess the Skill?? Aren't the events enough? (Exams, professions, use...)

Posted by Martin on 13/10/2019

Dear Franco,

I agree with all you say, but  I think you misunderstood me.

My approach is not documenting skills. My approach is documenting facts, rather than potentials. I take notice and may document that you spoke Latin, as I have done last time at school. I have a document stating my grade in Latin at high school.  My grade at high school confirms a set of years of continued successful lessons, not that I could understand much Latin now.
Speaking a language can be documented as an extended (observed) activity, as in FRBRoo. For instance, someone writing books in particular language. This falls under any kind of extended activity not further specified, such as an artist using a technique for some time, and avoids transforming actual activities into potentials.

We can document someone's documented opinion about a potential of a person, as an information object.

In the "Principles for Modelling Ontologies" we refer:
"7.2 Avoid concepts depending on a personal/ spectator perspective"

This could be elaborated more. In the CRM, we do not model concepts "because people use them", but because they can be used to integrated information related to them with URIs.  Therefore, your arguments and what I wanted to say is, "skill" is a bad concept for integration. What should be instantiated are the observable activities, which may or may not indicate skills.

The CRM does not make any statement that forbids describing concepts not in the CRM.

I kindly invite you to improve our principles document respectively

You write:
"So they are Actors even before doing anything or being in the position to do so, i.e. before any fact that can support assessing their capacity: just because the can “potentially" do it. Like myself with speaking Latin. Are they Actors since they are humans? Not really, because there is people “unfit to plead”, i.e. legally not responsible of their actions because of some mental infirmity. Nevertheless I would call them still Actors as they can do actions - without being responsible for what they do."

Of course.

This discourse applies to all concepts in the CRM. In dozens of tutorial we have stated, that all real-life concepts have fuzzy boundaries, and that scope notes to do not constitute sufficient logical conditions, but are sufficient to make users understand what we want to talk about. The argument you give is irrelevant, if the errors it may produce do not affect the recall of integrating relevant facts with the CRM. The CRM is not made to produce reliable AI results. Reality is not isomorphic to any logical model.  "Unfit to plead" does not change the fact that the kind of action is one a normal human can be made responsible for. How many "unfit to plead" appear in our cultural historical discourse, and would invalidate any of the properties we use? We indeed need to estimate if exceptions prevent us from finding relevant connections with CRM-based integration. For reasons of recall, we allow all concepts to include a few individuals unfit for the required properties.

You can continue if a chair with a broken leg is still a chair etc. If have not seen any concept free of fuzzy boundaries.

In the CRM, we do not define concepts as being the domain of a property, because then the primitive is the property.

I kindly invite you to improve our principles document respectively
 
to avoid such criticism at each concept, but much more to teach these principles effectively, or to provide better scope notes if there are any.

See also
"5.4 Model domains and range or properties consistent with your level of knowledge of the domain of discourse"

Posted by Martin on 14/9/2019

Dear George,

The first principle of all is are there relevant queries that need that property for integrating disparate sources, which indeed provide such data, and is that research one we like to support with the CRM?

Second, using p2 on E21 does the job, doesn't it? What is the added value of "knows language"?

Next principle, keep the ontology small. Querying 1000 properties is already more than anybody can keep in mind. Each additional property has an implementation cost. We need strong arguments for relevance.

It has been the mos t important success factor of the CRM to keep the ontology small and still expressive enough. If we loose this discipline, we will loose the whole project.

Finally, we are not repeating in the CRM the way typically information systems document, but always tried to find a more fundamental representation. With that argument, we could never have introduced events. They did NOT appear in any of the typical systems at that time. It is a principle not to model all the valuable description elements, which are relevant to characterize an item, but not creating interesting links across resources.

I did not say that it is a personal opinion that someone speaks a language. I said, this is observable. I document: Franco has spoken Latin, repeatedly? But talking about skills, is another level, it introduces a quality, which is hard to objectify, as Franco has pointed out. Actually, it is a typical classification problem, with all its boundary case questions, and the CRM is about relations between particulars.

So, what is the added value against p2, and what are the typical research data and typical research questions for integrating such data, that cannot be answered with P2?

Posted by George Bruseker on 14/10/2019

Dear Martin,

Well I am curious to know if I am the only one who finds this property useful at a high level for describing humans related to CH (presumably an important factor), or if I have truly not understood something.  So some comments below.
 

>     The first principle of all is are there relevant queries that need that property for integrating disparate sources, which indeed provide such data, and is that research one we like to support with the CRM?
>
>     Second, using p2 on E21 does the job, doesn't it? What is the added value of "knows language"? 

The has type solution has the following problem.

E21 p2 has type E56 Language does not work

It does not work because no E21 is a language. But we have already set off as a major category of documentation in the CRM an entire class for languages. It is called E56. So one would expect information around language to aggregate around... language or E56.

The solution: 

E21 p2 has type E55 English Speakers

is not satisfactory because you cannot make any connection between the type E55 English Speaker and any instance of language. You would have to model this relation, which would be an indirection, which would make a simple query people generally might want to make in a cultural heritage system difficult to make.

E21 knows language E56 http://vocab.getty.edu/page/aat/300388277 

would allow the user to connect some instance of E21 to E56 in a semantically clear way and that instance of E56 would be the same one they would use for talking about other objects in the graphs, thus aiding reasoning and helping in connecting things.

>
>
>     Next principle, keep the ontology small. Querying 1000 properties is already more than anybody can keep in mind. Each additional property has an implementation cost. We need strong arguments for relevance.
>
>     It has been the mos t important success factor of the CRM to keep the ontology small and still expressive enough. If we loose this discipline, we will loose the whole project.

In CRM Sig we decided to create a new extension CRMSoc in order to document social factors. I have proposed to put this new property here (though it is so general a type of information this could be argued). Presently we have as primary properties of actors 'contact point'. How, in a non arbitrary way, is this more or less general than language? It is a typical piece of information that one may know about an individual that can be useful in querying your data. Especially in a time of great interest in identity, knowing the variables that are pertinent to the actors related to CH collections does not seem very esoteric.
 

>
>     Finally, we are not repeating in the CRM the way typically information systems document, but always tried to find a more fundamental representation. With that argument, we could never have introduced events. They did NOT appear in any of the typical systems at that time. It is a principle not to model all the valuable description elements, which are relevant to characterize an item, but not creating interesting links across resources.

The proposal of 'knows language' is to follow yet another principle, model bottom up. So indeed, I would not rush to add 'has knowledge' because we don't know what that is, but we certainly do have a good enough mesoscopic understanding of what it means for someone to know a language (in the same rough and ready way we speak of somethign consisting of a material). Of course this could be specialized for highly specific anthropological studies, but this does not fall into a use case I am aware of. Almost any good CH database I have modelled that has information about the people related to the collection does document this. 

I agree we can't add any property for our pleasure, but this is really fundamental documentation, not at all on the sidelines of the information space we are talking about. It is equivalent to saying an object has a material. The people doing CH fall in scope and that they know a language is as fundamental as that. 

About the structure of building the model, of course we should not have it reflect the systems. The principle I refer to is the principle that we have to model at the level of knowledge that is typically available. That is why CRM has shortcuts and allows multiple levels of description given what is known by the user. So we do not force everything to go through an event when we know that typically such thing will not be known (but may in principle be). So it would be great to imagine also the 'language acquistion' node which is the moment(s?) of acquiring a language which would pertain to a very specific field of study. But for the CH information system, we do know that somebody knew a langauge and it is valuable to the overall study and so, unclear to me why not necessary. Introducing 'knows language' introduces interesting links between resources...see above.
 

>
>     I did not say that it is a personal opinion that someone speaks a language. I said, this is observable. I document: Franco has spoken Latin, repeatedly? But talking about skills, is another level, it introduces a quality, which is hard to objectify, as Franco has pointed out. Actually, it is a typical classification problem, with all its boundary case questions, and the CRM is about relations between particulars.
>
>     So, what is the added value against p2, and what are the typical research data and typical research questions for integrating such data, that cannot be answered with P2?

It would be difficult for me to generate all the potential research questions that people may want to do based on knowing the languages that people related to cultural heritage objets through events might want to perform. Again, identity studies are very large, today. 

You may want to know the number of English knowers vs Greek knowers in digs at different periods in the history of archaeological research. You may want to try to see patterns in collections related to speakers of X as opposed to speakers of Y language. You may want to know who at a certain time in your graph may have been able to read a book from that graph. 

I struggle to see the esoteric ness  of the proposal and do see how it fits to the modelling principles we elaborated.

Posted by Detlev Balzer on 14/10/2019

Dear George, Martin,

this discussion made me curious whether or not I can confirm George's assertion that such statements are common in the cultural heritage field.

EAC-CPF does have a language element, which is, however, only used to indicate in which language the name of a person or corporation is expressed. 

GND, the authority file for libraries in German-speaking countries, has a Language entity which is used for making statements about the "field of study" of a person. Other predicates for the person-language pair of entities do occur, but these are obvious data entry errors.

Having extracted person-related data from a dozen or more cultural heritage projects, I don't remember any example where languages spoken or known by somebody have been considered in any other sense than relating to the documented activity, rather than to the (possibly un-instantiated) capacity of the person.

Of course, this is just an observation that doesn't prove anything. Personally, I would tend towards Martin's view that there is little, if anything, to be gained by defining such kind of statement in a reference model such as the CIDOC CRM.
 

Posted by George Bruseker on 15/10/2019

Dear all,

I think that the turn to the data is the right move here.

So some examples at hand:

Viaf - widely used ref

http://viaf.org/viaf/27251336

Emu - widely adopted collections management system

Chin makers in Canada model - national standard body Canada

It is in the requirements by researchers for building their persons model

Finally, I would come back to the initial example. The project in question as far as I follow it, looks to extract analytic data typically encoded by researchers in archival formats that HAVE NOT ALLOWED the precise recording of information that they would have liked to record Analytically and formally. Ie they are precisely trying to find good structures where there were none. The solution is certainly not to document it as an information object.

I am glad to keep finding examples. I hope we can use this practice generally and could even think of formalizing a list of schemas of reference.

Best,

Posted by Pat on 7/03/2021

In conclusion: using F51 Pursuit as currently defined would cover any professional/creative contexts of language use; covering personal language use would require broadening the scope note of F51 to include any specific period of continuous activity.

Post by George Bruseker (27 January 2022)

Hi all,

Following up on this issue, I have had a hack at proposing a property name and scope for a general skills type property. Here it is:

https://docs.google.com/document/d/1kge1gqohcr4NQK89kxbhiLfGuqgzassJpWze...

Best,

George
 

Post by George Bruseker (27 January 2022)

Hi Matthew,

There is a long trail on this discussion nwhere there is quite some disagreement whether we can assert knowledge or skill regarding someone so I was going for some phrasing that was as light as possible and hence landed on aptitude. I would be happy with has knowledge of or has skill in or has know how of, but aptitude was an effort to be in line with epistemic conservatives. Happy to strengthen it if everyone else would want.

Welcome back! 

Best,

George

In the 57th CIDOC CRM & 50th FRBR/LRMoo SIG Meeting, the SIG was informed about the latest developments on the issue. 

A group (GB, PR, TV, MR) had formed to work on an “ability/aptitude” property in May 2022. They have not shared any HW with the SIG since, namely: examples pointing to the relevance for the scope of CRM. 
The SIG resolved to not close the issue for the moment, since none of the group were present at the discussion. The group will be given some more time (until the spring 2024 meeting in Paris) to prepare the HW. If no HW is drafted by then, the issue will close. If there is HW available at the time, the SIG will decide accordingly.

Marseille, October 2023