- From: Phil Barker <phil.barker@pjjk.co.uk>
- Date: Wed, 24 Jan 2018 13:24:17 +0000
- To: public-eocred-schema@w3.org
- Message-ID: <355bd606-2e70-2359-db87-5fc2f128a7ea@pjjk.co.uk>
Hello Fritz,
On 23/01/18 22:44, Fritz Ray wrote:
> Hi Phil,
>
> I appreciate the pragmatic approach, and I don't disagree with
> anything you have said.
>
> Where I don't want to get is where we're at with
> schema.org/CreativeWork <http://schema.org/CreativeWork> with
> "learningResourceType", defined as being able to be things like
> "presentation" or "handout" -- These are more than just labels, these
> are indicators of what properties the resource has, and violate the
> basis of object oriented structures, where 'is a' relationships are
> defined by subclasses and 'has a' relationships are described in
> properties.
I don't want to get too involved in wider-scale modelling within
schema.org (at least not here), but learningResourceType is quite an old
bit of schema.org, which predates any concepts like DefinedTerm. I think
that as soon as DefinedTerm is secure in schema.org we will be wanting
to add it to the expected range of learningResourceType. There similar
issues with the number different schemes for learningResourceType (and
throw in some vastly different conceptualization of 'type').
>
> Consider which sounds more clear:
> Presentation is a creative work, Handout is a creative work.
> Creative work has a learning resource type of Presentation.
>
I have no problem with people doing both. There will always be some
types which aren't represented in the schema.org hierarchy, and for
which there is no suitable URI linking to some other vocabulary: how
will these be represented if there is no learningResourceType?
> What happens is that at some point down the line, these differences
> matter enough to want to functionally distinguish between, say, a
> Credential and a Qualification, or in the above, a presentation and a
> handout. Presentation (as a creative work) may have a slide count, but
> handouts don't. Do I add a slide count property to CreativeWork? What
> happens when I create Presentation as a subtype of CreativeWork, and
> what does that mean for apps?
You may find apps consuming schema.org have to do more work (in this
case looking at learningResourceType and the schema.org class), while it
is easier for publishers of data to publish about resources for which
there is no specific schema.org class.
>
> This has exactly happened. PresentationDigitalDocument is a type of
> DigitalDocument is a type of CreativeWork, and
> http://schema.org/subEvent states that a Presentation is probably best
> subtyped from Event.
[aside: note the problem of Presentation being a CreativeWork which is
presented (e.g. a powerpoint deck) or the event at which it is presented]
>
> So, do I create a CreativeWork with learning resource type
> Presentation, or do I create a PresentationDigitalDocument? Do I
> redundantly label the PresentationDigitalDocument with learning
> resource type: Presentation?
>
Personally, I would create PresentationDigitalDocument with redundant
resource type.
> I get wanting to allow people to describe things in their own way, but
> subtypes described as properties create enormous problems, and just as
> http://schema.org/Course doesn't allow itself to be labelled with
> Course, Class, Lecture, Workshop, Lab or any of the other variants
> (IgniteTalk, MOOC),
Modelling again: They may be considered learning resource types, or (my
preference) the same course could be presented as an online or a face to
face class. CourseInstance <http://schema.org/CourseInstance> does have
courseMode <http://schema.org/courseMode> property for exactly this.
> I don't think we should either.
So, are you saying that we should strike the use case of wanting to
provide information about subtypes of credential? If not, how would you
encode SQA's HNCs <https://www.sqa.org.uk/sqa/168.2432.html> as a
subtype of credential?
> Especially in light that many of these different types of credentials
> will have additional properties, and will warrant subtypes. Degrees
> occur at masters, bachelors, associates levels, but qualifications don't.
In the UK, degrees are commonly called qualifications. See also first
sentence of wikipedia:Academic degree
<https://en.wikipedia.org/wiki/Academic_degree>. So are HNCs for that
matter. Did I mention that agreeing on a typology for global use would
be difficult?
> Specific classes of teaching certificates may only qualify at
> particular grade levels, and that's something you want in the subtype
> but not in EOC.
I think the specifics of that goes beyond our current use cases. We do
have a use case around finding jobs / occupations that match a
credential (or vice versa), so if Primary School Teacher is an
occupation distinct from Teacher the specific EOCredential required for
that should be findable.
Perhaps I am still not getting your proposal. If so, could you show some
specific examples or how it would work, say with SQA HNCs?
Phil
>
> On Thu, Jan 18, 2018 at 3:42 AM, Phil Barker <phil.barker@pjjk.co.uk
> <mailto:phil.barker@pjjk.co.uk>> wrote:
>
> Hello Fritz, I don't disagree with many of the points you raise, I
> think you are (mostly) factually correct, but I do disagree with
> your conclusion. This based on pragmatic judgement of what we can
> create and maintain, what we can expect to be accepted as
> additions to schema.org <http://schema.org>, and what we can
> expect providers of data to do.
>
> additionalType is a possibility, worth discussing separately. My
> concern is that it lacks support for text values. You and I both
> wish that more data were provided as URIs, but the reality is that
> there are no URIs for many of the credential subtypes. We have to
> support text values.
>
> Adding subtypes of EducationalOccupationalCredential to schema.org
> <http://schema.org> for all the specific types of credential is a
> non-starter for me. There would be thousands. We couldn't hope to
> collect them all, and schema.org <http://schema.org> would not
> accept so many.
>
> Creating a typology along the lines of what is in CTDL, I also do
> not think would work. I think 20 or so sub types would be too many
> for schema. Doing anything meaningful by way of a global
> classification of credential types is not a task I would want to
> take on (I've seen how difficult it is just within Europe). Some
> examples of the difficulties can be seen in trying to fit European
> credentials to CTDL. The distinction between Diploma and
> Certificate in the UK's HND and HNC are not well conveyed by the
> definitions for ceterms:Diploma and ceterms:Certificate, and
> losing distinctions like this would be detrimental to fulfilling
> the use case. I can also imagine all sorts of errors in data
> provision arising from instances like the old-style German Diplom
> and Spanish Licenciatura (comparable to Masters degrees, not
> diploma or licence).
>
> I do not agree that pointing to an externally defined term
> discourages subtyping. If a DefinedTerm is part of a
> DefinedTermSet, relationships between terms may be defined using
> vocabularies such as SKOS. However, the harsh reality in which we
> have to operate is that they are more likely to be defined in
> pdfs. Sorry.
>
> Finally, remember the use case is to allow people to search for a
> specific type of credential. For example, "where can I find
> information about PGCEs?". credentialType="PGCE" would suffice in
> many cases.
>
> Phil
>
>
> On 17/01/18 18:07, Fritz Ray wrote:
>> I want to say that Stuart and I have already had this discussion,
>> but:
>>
>> I don't think credentialType is needed.
>>
>> Adding credentialType conflicts with the intent of the object's
>> type. Remember, type is just a field as well.
>>
>> Schema.org/Thing also has http://schema.org/additionalType
>> <http://schema.org/additionalType> to permit the adding of
>> additional (specific, not defined here) type data to any object.
>>
>> http://schema.org/Action is a good example of where a lot of
>> subTypes are added to a base class without necessarily adding
>> properties (though some do!), so we also have precident there.
>>
>> There's lots of established avenues that allow for subtyping. On
>> the other hand, there are also several instances of additional
>> type being specifically identified, usually paired with a fixed
>> taxonomy of types (ActionStatusType, BoardingPolicyType,
>> BusinessEntityType, EventStatusType). I also believe these
>> instances aren't in the spirit of schema.org <http://schema.org>.
>>
>> Type properties like credentialType prevent additional subtyping
>> and encourage "text" descriptions (see AlignmentObject's
>> AlignmentType) where it should really be URLs so that
>> interpretations can be distinguished and separate.
>>
>> Type properties also discourage additional subtyping and do not
>> prevent a means of structuring dependent subtypes. Is an IVA
>> credential (from the example above) a type of SVQ credential?
>> Without an additional structured ontology to point at, it is
>> difficult to know.
>>
>> What additional properties does a specific type of credential
>> confer? For instance, a military MOS credential may have a
>> securityClassificationLevel. Do we need to add those properties
>> to the base class? (With credentialType, maybe, with subtypes,
>> certainly not.)
>>
>> On Wed, Jan 17, 2018 at 9:04 AM, Stuart Sutton
>> <stuartasutton@gmail.com <mailto:stuartasutton@gmail.com>> wrote:
>>
>> Phil, I agree that a credentialType property is needed. I
>> think that having the range include both Text _and_ URL to
>> allow identification by URI where available and text where
>> not is appropriate. You are right that CTDL defines
>> subclasses of it's Credential class
>> <http://purl.org/ctdl/terms/Credential> and that no single
>> enumeration will handle the breadth of the range (and
>> internationalization issues); but, where such enumeration(s)
>> exist, my sense is that they should be used. The CTDL defines
>> a credentialType <http://purl.org/ctdl/terms/credentialType>
>> property that is intended to be used wherever reference to a
>> member of the Credential class is needed. It's range now
>> enumerates the CTDL subclasses; but should likely reference
>> more inclusively the Credential class.
>>
>> As for use of TermDefinition to identify/define members of
>> such enumerations, I hope that it gets approved in a timely
>> manner and that we are not forced to even contemplate use of
>> AlignmentObject (ugh...don't get me started).
>>
>> Stuart
>>
>> On Wed, Jan 17, 2018 at 3:20 AM, Phil Barker
>> <phil.barker@pjjk.co.uk <mailto:phil.barker@pjjk.co.uk>> wrote:
>>
>> Hello again, moving on to the next requirement for
>> describing Educational and Occupational Credentials in
>> schema.org <http://schema.org>: I suggest we look at how
>> to identify the subtypes of these credentials.
>>
>> The use case for this
>> <https://www.w3.org/community/eocred-schema/wiki/Use_Cases#Identify_subtypes_of_credential>
>> gives examples of "degree" "certificate" "badge". I know
>> there are about 20 others from the Credential Engines'
>> CTDL <http://credreg.net/ctdl/handbook#creds>. Most
>> countries will have their own types of EO Credential, for
>> example in Scotland we have National Qualifications,
>> HNDs, HNCs, SVQs, IVAs, PDAs, DipHEs, CertHEs and many
>> more. Other countries will be similar. Furthermore, the
>> types of qualification on offer changes over time.
>>
>> In short, the number of types is we need to consider is
>> vast and varied. So, while CTDL has subclasses of its
>> Credential class for each of its distinct types, that is
>> not a practical solution for wider use. Even if we could
>> reduce the number and variety of types, I think it would
>> add too many subclasses to the schema.org
>> <http://schema.org> hierarchy, given that most of the
>> subtypes would have no unique properties.
>>
>> The alternative is for EducationalOccupationalCredential
>> to have a property which records the type of credential.
>> With a nod to Richard's point that much of what we do is
>> applicable to generic credentials, I propose we call this
>> credentialType.
>>
>> The basic range for credentialType would be text, and I
>> think we should explicitly allow this. We could stop here.
>>
>> In an ideal world there would be controlled vocabulary
>> for naming the credentialTypes. However, I a single
>> controlled vocabulary of all the precise types is not
>> feasible, and I think that producing a vocabulary that
>> classifies these types into categories like "certificate"
>> would be very difficult and the results would be very
>> imprecise. We should, however try to facilitate the use
>> of local controlled vocabularies. This is where we reach
>> the edge of what currently possible in schema.org
>> <http://schema.org>.
>>
>> Options for facilitating the use of local controlled
>> vocabularies of credential type:
>>
>> 1, allow a URL to link to a controlled value / external
>> enumeration.
>>
>> 2, allow alignmentObjects to provide information about
>> the credentialType as if credential types were
>> educational frameworks
>>
>> 3, use the developing schema.org <http://schema.org> type
>> that is currently called CategoryCode
>> <http://pending.schema.org/CategoryCode>, but which is
>> proposed to be changed to TermDefinition
>> <https://github.com/schemaorg/schemaorg/issues/1775>
>>
>> In my view: 1 is too vague (who knows what will be at the
>> end of the URL), 2 stretches the alignmentObject
>> somewhat, and 3 is the best option for the long run. An
>> example using option 3 would look something like:
>>
>> {
>> "@type": "EducationalOccupationalCredential",
>> "name" : "HNC Facilities Management",
>> "credentialType": {
>> "@type" : "TermDefinition",
>> "name" : "Higher National Certificate",
>> "termCode" : "HNC",
>> "inDefinedTermSet" : "SQA Qualifications" //should be
>> a URL or DefinedTermSet object
>> }
>> }
>>
>> What do you think? Too complicated, maybe? Am I
>> overthinking the problem? Are there enough
>> well-constructed sets of terms describing credential
>> types for it to be worth trying to accommodate anything
>> other than text values?
>>
>> Phil
>>
>> --
>>
>> Phil Barker <http://people.pjjk.net/phil>.
>> http://people.pjjk.net/phil
>> PJJK Limited <https://www.pjjk.co.uk>: technology to
>> enhance learning; information systems for education.
>> CETIS LLP: a cooperative consultancy for innovation in
>> education technology.
>>
>> PJJK Limited is registered in Scotland as a private
>> limited company, number SC569282.
>> CETIS is a co-operative limited liability partnership,
>> registered in England number OC399090
>>
>>
>>
>>
>> --
>> Stuart A. Sutton, Metadata Consultant
>> Associate Professor Emeritus, University of Washington
>> Information School
>> Email: stuartasutton@gmail.com <mailto:stuartasutton@gmail.com>
>> Skype: sasutton
>>
>>
>>
>
> --
>
> Phil Barker <http://people.pjjk.net/phil>. http://people.pjjk.net/phil
> PJJK Limited <https://www.pjjk.co.uk>: technology to enhance
> learning; information systems for education.
> CETIS LLP: a cooperative consultancy for innovation in education
> technology.
>
> PJJK Limited is registered in Scotland as a private limited
> company, number SC569282.
> CETIS is a co-operative limited liability partnership, registered
> in England number OC399090
>
>
--
Phil Barker <http://people.pjjk.net/phil>. http://people.pjjk.net/phil
PJJK Limited <https://www.pjjk.co.uk>: technology to enhance learning;
information systems for education.
CETIS LLP: a cooperative consultancy for innovation in education technology.
PJJK Limited is registered in Scotland as a private limited company,
number SC569282.
CETIS is a co-operative limited liability partnership, registered in
England number OC399090
Received on Wednesday, 24 January 2018 13:24:48 UTC