Why no "LearningResource" type?

[Paul Libbrecht]

Hello LRMI fellows,

I hope this mailing list is still in use.
Otherwise, I'm fine to join the W3C group if useful (it sounded to be
focused to courses).

We've been attempting to extend's Phil Barker's attempt to use Google
Custom Search Engine with LRMI schema.org markup and bump into an
elementary wall: apparently google custom search engine only allows to
filter by schema.org types and not by properties (values or simply
existence of a property statement).

Hence, thus far, we can only search for EducationalAudience but we would
like to search for any LearningResource which I would expect to be a
sub-type of CreativeWorks.

Did I miss something?

Paul

PS: when I land on LRMI.net, I only see the Knowledge-Base from the
carrousel. Could it go in the menu please?

[Phil Barker]

Hello Paul, this is exactly the right list.

We took a deliberate decision way back at the start of LRMI not to define a LearningResource as a subtype of CreativeWork. Essentially the problem comes when you try to define what is a Learning Resource. Everyone who has tried so far has come up with something like "a resource which is used in learning, education or training". That doesn't rule out anything. So rather than rehash the problem of deciding what is and isn't a learning resource, we took the approach of providing a way by which people can describe the educational properties of any Creative Work.

We recognised that there are some "types" of resource that are specific for learning (textbooks are qualitatively different to novels, instructional videos are not like the movies people would watch in the cinema, even though novels and movies are useful in education), that's why we have the learningResourceType property. You can think of this as describing the educational genre of the resource.

In practice there are two choices for searching for learning resources. You can search those sites that are curated collections of what someone has decided are educational resources. Or you can search for the educational properties you want. So in our attempt at creating a Google Custom Search Engine we looked for the AlignmentObject. Looking for the presence of a learningResourceType would be another way. The educationalUse property should likewise be a good indicator.

Your student's project looks really interesting. I had a message from him yesterday which I haven't had time yet to answer (I will). I hope discussing this question here raises awareness of the project.

And yes, the LRMI.net site needs some attention. I believe it is in hand, but I'll make sure your comment gets to the right people.

All the best, Phil

--
You received this message because you are subscribed to the Google Groups "Learning Resource Metadata Initiative" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lrmi+uns...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

-- 
--  
Phil Barker           @philbarker
LRMI, Cetis, ICBL     http://people.pjjk.net/phil
Heriot-Watt University

Ubuntu: http://xkcd.com/456/
  not so much an operating system as a learning opportunity.

[Paul Libbrecht]

Phil Barker wrote:
> That doesn't rule out anything. So rather than rehash the problem of
> deciding what is and isn't a learning resource, we took the approach
> of providing a way by which people can describe the educational
> properties of any Creative Work.

but that means that there's no way to simply say "this is a learning
resource".
I sure do not want to have a formal definition of that, just a way to
say it ! ;-)

paul

[Phil Barker]

I feel your pain, but I don't think there can be an simple way to say
that something is a learning resource until you can define whether
something is a learning resource. The presence of a property like
learningResourceType is your best bet at inferring that someone thinks a
resource is a learning resource. That's when you start to realise that
what some people would class as a learning resource does not necessarily
match with what you would expect to find classed as a learning resource
(at least, that was the case for me--when I saw lifestyle magazines
classed as learning resources--it was language learning--I decided to
give up any attempt at treating learning resource as distinct from
creative work).

As an aside, while it would be possible to say a resource is a
LearningResource and a Video, I don't know how well tools support
classification of a resource as more than one schema type so that they
can be described with properties from both.

Phil

[Simon Grant]

On 22 January 2016 at 11:33, Phil Barker <phil....@hw.ac.uk> wrote:

On 22/01/16 11:14, Paul Libbrecht wrote:

Phil Barker wrote:

That doesn't rule out anything. So rather than rehash the problem of
deciding what is and isn't a learning resource, we took the approach
of providing a way by which people can describe the educational
properties of any Creative Work.

but that means that there's no way to simply say "this is a learning
resource".
I sure do not want to have a formal definition of that, just a way to
say it ! ;-)

I feel your pain, but I don't think there can be an simple way to say that something is a learning resource until you can define whether something is a learning resource. The presence of a property like learningResourceType is your best bet at inferring that someone thinks a resource is a learning resource. That's when you start to realise that what some people would class as a learning resource does not necessarily match with what you would expect to find classed as a learning resource (at least, that was the case for me--when I saw lifestyle magazines classed as learning resources--it was language learning--I decided to give up any attempt at treating learning resource as distinct from creative work).

As an aside, while it would be possible to say a resource is a LearningResource and a Video, I don't know how well tools support classification of a resource as more than one schema type so that they can be described with properties from both

I'm following Phil here ... but maybe another way of thinking about this is to treat "learning resource" as a relationship or property rather than a class? In which case it might be called rather something like "has learning resource". That would definitely make sense to me, by way of saying that a resource is a resource for a course (of whatever granularity) of learning, education, or training. Alternatively, a resource could be one of "my" learning resources, in which can it would be "me" who "has learning resource", meaning perhaps something like "I have found this useful in my learning". Or even, one resource could "have" another learning resource, meaning that the learning resource was helpful in learning about the first resource. Again, a property or a relationship.

I haven't checked whether or not LRMI facilitates these kinds of statement.

Simon

--

from Simon Grant +44 7710031657 http://www.simongrant.org/home.html

[Joshua Marks]

Paul,

I have always considered that any page that contains at least one LRMI property for CreativeWork can be inferred as being usable as a learning resource (In the context of the supplied metadata at least.) So using Google Custom Search you only need to filter for pages with any LRMI property then you can filtered further by LearnignResourceType if that exists. 

Regarding "My" learning resource, all you need to do is add a web link to it in Curriki or similar site that captures metadata and tags the related Schema.org values, and there you go, it is your learning resource already tagged with your context. 

I hope you are doing well and that that helps. 

-Joshua

--

[Paul Libbrecht]

Hey Joshua,

> I have always considered that any page that contains at least one LRMI
> property for CreativeWork can be inferred as being usable as a
> learning resource (In the context of the supplied metadata at least.)
> So using Google Custom Search you only need to filter for pages with
> any LRMI property then you can filtered further by
> LearnignResourceType if that exists.

I would definitely agree. Just that Google Custom Search Engine does not
allow that (apparently, question asked on stackoverflow, I would love
hearing the contrary).

> Regarding "My" learning resource, all you need to do is add a web link
> to it in Curriki or similar site that captures metadata and tags the
> related Schema.org values, and there you go, it is your learning
> resource already tagged with your context.

Then the-real-place for my learning resource would be Curriki, or?
I am more and more bothered by links...
Ideally should Google CSE consider a Curriki link page as an additional
metadata to the original resource, or?
There's surely a danger doing that but I feel this would be the best option.

Paul

[Phil Barker]

On 22/01/16 15:52, Paul Libbrecht wrote:

Hey Joshua,

I have always considered that any page that contains at least one LRMI
property for CreativeWork can be inferred as being usable as a
learning resource (In the context of the supplied metadata at least.)
So using Google Custom Search you only need to filter for pages with
any LRMI property then you can filtered further by
LearnignResourceType if that exists.

I would definitely agree. Just that Google Custom Search Engine does not
allow that (apparently, question asked  on stackoverflow, I would love
hearing the contrary).

AFAIK, you can customise by limiting the search engine as a whole to some URL pattern or pages that contain some schema.org type (which I guess is what the stackoverflow question relates to) but you can also create refinements that filter by schema.org property value. A query
more:p:videoobject-accessibilityfeature:captions
will prioritise pages which have a VideoObject tagged with accessibilityFeature = captions

(Look under for the refinements tab in search features in the GCSE setup, see slides 9, 10 & 11 at
http://www.slideshare.net/philb/lrmi-implementation-by-google-custom-search )

We found that giving priority to site with this label worked better than searching only for sites with this label.

Phil

Regarding "My" learning resource, all you need to do is add a web link
to it in Curriki or similar site that captures metadata and tags the
related Schema.org values, and there you go, it is your learning
resource already tagged with your context.

Then the-real-place for my learning resource would be Curriki, or?
I am more and more bothered by links...
Ideally should Google CSE consider a Curriki link page as an additional
metadata to the original resource, or?
There's surely a danger doing that but I feel this would be the best option.

Paul

[Stuart Sutton]

I have to say that in most things having to do with metadata for education/training resources, Phil and I usually see eye-to-eye or come arguably close; but, we somewhat part company with this issue of whether there should be an explicit LearningResource subtype of CreativeWork. There is an open issue on the LRMI github proposing that LRMI pursue such a subtype [1].  Phil argues against it there as well.

Phil is absolutely correct that if you ask those people who are supposedly in the "educational know" for a definition of a learning resource class you get a typically rich array of conflicting notions. On the fringes of those opinions are people to asset that all things are learning resources--just watch a child learning from its world of all things if you don't believe it. But all that leads nowhere or to no class at all for lack of distinguishing characteristics. But, suggest that there is no such things as a learning resource to a K-12 teacher and she may let you know that in her world, you are peddling fiction. In the end, Paul says he doesn't care about how you formally define it, he wants to mention it's name and get a result.

Phil actually does outline a very pragmatic definition of a LearningResource type with which I can completely agree.  Here's what Phil appears to say. We can infer a LearningResource type (even if we cannot speak its name) from the presence of certain properties that announce the educational intentions of the resource (his example, use of schema.org/learningResourceType). So one prong of a such a definition is that the candidate resource announces it's educational intention through use of one or more LRMI properties. Second, Phil has given up trying to treat the notion of a "learning resource as distinct from CreativeWork". I'd say that is good since their is no inherent difference that I can discern.

So it appears that Phil and I agree that a CreativeWork that announces its educational intentions can be inferred to be a LearningResource. Where we disagree is that I'd go a step further and make the inference explicit through the addition of a LearningResource subtype of CreativeWork. Why bother? Because then we can stop causing difficulties for people like Paul who want to (1) find it (not figure out how to infer it); and (2) state it's name. And there are probably more than a few who, unlike Paul, don't say anything and assume there is no way to markup such a beast using schema.org.

Stuart

[1] https://github.com/dcmi/LRMI-Terms/issues/11

[Simon Grant]

Has anyone considered having a property "has learning resource" rather than a class "learning resource"?

I imagine its domain would be any resource, and its range would be a creative work.

The inverse property might be called "is learning resource for" and would have domain creative work and range any resource.

For all I know, it may have been considered and rejected for good reasons.

But still, at this point maybe an answer to that would shed a little more light?

Thanks

Simon

[Phil Barker]

Hello Simon

On 23/01/2016 04:39, Simon Grant wrote:

Has anyone considered having a property "has learning resource" rather than a class "learning resource"?

I imagine its domain would be any resource, and its range would be a creative work.

This would be a property of Thing (since you can learn about any Thing) pointing to those resources that would help you learn about that Thing. I am not sure how this would work out in practice, but it would perhaps be very similar to http://schema.org/mainEntityOfPage because if the main entity being described on a page is Cath's Cafe then that page is a good resource to use to learn about Cath's cafe.

The inverse property might be called "is learning resource for" and would have domain creative work and range any resource.

The http://schema.org/educationalAlignment is essentially this inverse property. You can you it and the http://schema.org/AlignmentObject to say "this resource teaches competence X" or "this resource is about academic subject Y". The difference to what you suggest is that it describes alignment to things in educational frameworks, so the range would something like a concept scheme, however there is nothing like a concept scheme in schema.org hence the use of targetName, targetDescription, targetUrl in the AlignmentObject.

Hope this helps, Phil

-- 
Phil Barker           @philbarker
LRMI, Cetis, ICBL     http://people.pjjk.net/phil

Heriot-Watt University

Workflow: http://www.icbl.hw.ac.uk/~philb/workflow/

We invite research leaders and ambitious early career researchers to join us in leading and driving research in key inter-disciplinary themes. Please see www.hw.ac.uk/researchleaders for further information and how to apply.

Heriot-Watt University is a Scottish charity registered under charity number SC000278.

[Simon Grant]

Thanks, Phil, that is at least a start to understanding where we are with this.

Am I alone in having difficulties with AlignmentObject? Maybe it's because I wasn't in on the development of AlignmentObject, but to me it seems quite abstract and generic, and I guess it must confuse more people than just me!  I'm sure LRMI had good reasons for not choosing something simpler and easier to understand. But I do wonder what they are...

My sense is that a property "is learning resource for", with or without "has learning resource", would be far simpler and easier to understand than EducationalAlignment together with AlignmentObject, and therefore not only much more likely to be adopted and used in the longer term, but also more likely to result in higher quality usable microdata, because, being simpler, people are more likely to use it consistently.

Just saying.

Simon

--

You received this message because you are subscribed to the Google Groups "Learning Resource Metadata Initiative" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lrmi+uns...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Simon Grant]

Hi again Phil and all

I'd like to elaborate on this a little, if I may...

On 25 January 2016 at 09:42, Phil Barker <phil....@hw.ac.uk> wrote:

Hello Simon

On 23/01/2016 04:39, Simon Grant wrote:

Has anyone considered having a property "has learning resource" rather than a class "learning resource"?

I imagine its domain would be any resource, and its range would be a creative work.

This would be a property of Thing (since you can learn about any Thing) pointing to those resources that would help you learn about that Thing. I am not sure how this would work out in practice, but it would perhaps be very similar to http://schema.org/mainEntityOfPage because if the main entity being described on a page is Cath's Cafe then that page is a good resource to use to learn about Cath's cafe.

Yes, if a CreativeWork is mainly about an Entity, then it will most likely be a learning resource for that entity. Of course, many learning resources are about several Things, and as I read it, it wouldn't be correct to say that several Things have mainEntityOfPage as the same CreativeWork. Thus, mainEntityOfPage wouldn't be an effective substitute for "has learning resource".

I was going to write "hope this helps" here, but I have an uncomfortable feeling that my comments here may be unwelcome, in that I'm suggesting something that doesn't fit in with the way people have settled on for the time being to do things. Sorry about that! But I do hope that it is worth raising this, if it might result in a better solution for everyone in the longer term.

Thanks

Simon

[Phil Barker]

On 25/01/2016 10:20, Simon Grant wrote:

Thanks, Phil, that is at least a start to understanding where we are with this.

Am I alone in having difficulties with AlignmentObject? Maybe it's because I wasn't in on the development of AlignmentObject, but to me it seems quite abstract and generic, and I guess it must confuse more people than just me!  I'm sure LRMI had good reasons for not choosing something simpler and easier to understand.

'Yes' to all points.

But I do wonder what they are...

This was my best shot at explaining what the alignment object is, with some explanation of how it got to be that way and why: http://blogs.pjjk.net/phil/explaining-the-lrmi-alignment-object/

Broadly speaking there are two reasons behind the AlignmentObject
(1) an extra level of indirection, which is useful when you don't want to specify every possible type of alignment in the properties of the thing that is aligned from or when you want to reify the alignment relationship in order to say some things about it (such as who says this resource is good for teaching that competency); and
(2) a surrogate for the thing that is being aligned to,  which is useful when there is no authoritative machine readable representation of the thing being aligned to.

Phil

[Simon Grant]

On 25 January 2016 at 10:34, Phil Barker <phil....@hw.ac.uk> wrote:

[...] This was my best shot at explaining what the alignment object is, with some explanation of how it got to be that way and why: http://blogs.pjjk.net/phil/explaining-the-lrmi-alignment-object/

Ah, I read Phil's blog and even commented on it shortly after it was written! So I can claim only forgetfulness, not ignorance ;)
 

Broadly speaking there are two reasons behind the AlignmentObject
(1) an extra level of indirection, which is useful when you don't want to specify every possible type of alignment in the properties of the thing that is aligned from or when you want to reify the alignment relationship in order to say some things about it (such as who says this resource is good for teaching that competency); and
(2) a surrogate for the thing that is being aligned to,  which is useful when there is no authoritative machine readable representation of the thing being aligned to.

And I guess we're mostly holding in our minds (in tension) *both* the fact that there are use cases which make AlignmentObject useful, *and* the fact that people seem easily confused by the double abstraction of EducationalAlignment together with AlignmentObject.

To me, it would seem that the ability to make either one of these abstractions concrete, rather than abstract, would make life easier for people trying to support or use LRMI in these cases.

1. As a parallel alternative (*not* a replacement) one could perhaps allow people to use a very small number of the most useful types of educationalAlignment, essentially taking recommended values of the alignmentType from the AlignmentObject. Sure, this would make little or no difference to the logic, just to the ease of understanding and use. Maybe it would be easier for people to see that something is a learning resource in virtue of having a straightforward "teaches" property, instead of an educationalAlignment to an AlignmentObject where the alignmentType is "teaches".

On the other hand, maybe aligning to an AlignmentObject doesn't capture what is being looked for. AlignmentObject points to an (node in an) educational framework, not a Thing in general, and it looks like people could use something that relates to a Thing.

2. If we had some of these concretised educationalAlignment properties, we could allow suitable ones to have Things as their ranges, rather than AlignmentObjects? Sure, at present it wouldn't seem to make any sense to have an educationalAlignment directly to a Thing, but maybe that's because the type of alignment has been packed away in the AlignmentObject, instead of being an integral part of the educationalAlignment itself. And my understanding is that, well, it couldn't be...

Really hoping that this is more helpful than counterproductive, and with apologies in advance if people find the latter.

Simon

[Stuart Sutton]

On Mon, Jan 25, 2016 at 2:34 AM, Phil Barker <phil....@hw.ac.uk> wrote:

This was my best shot at explaining what the alignment object is, with some explanation of how it got to be that way and why: http://blogs.pjjk.net/phil/explaining-the-lrmi-alignment-object/

Broadly speaking there are two reasons behind the AlignmentObject
(1) an extra level of indirection, which is useful when you don't want to specify every possible type of alignment in the properties of the thing that is aligned from or when you want to reify the alignment relationship in order to say some things about it (such as who says this resource is good for teaching that competency); and

My recollection with regard to the first reason (to avoid explicitly "specifying every possible type of alignment" as a property) is spot on with yours, Phil. As is the second, reification of the relationship  so you can comment on the relationship itself (e.g., who asserts the alignment, strength of the alignment, etc.). The latter is a function with considerable potential.

 

(2) a surrogate for the thing that is being aligned to,  which is useful when there is no authoritative machine readable representation of the thing being aligned to.

This is the necessary (but somewhat unfortunate) use that can lead to the AlignmentObject being used as a simple substitute for a description of the object of the relationship. This creates even more confusion on top of the already confusing indirection noted by you, Phil, since, in RDF terms (and schema.org is an RDF model), an instance of the AlighmentObject becomes both a refication of the relationship (i.e., the assertion's predicate)  and the assertions object mushed together.

[Stuart Sutton]

On Mon, Jan 25, 2016 at 3:19 AM, Simon Grant <asi...@gmail.com> wrote:

On 25 January 2016 at 10:34, Phil Barker <phil....@hw.ac.uk> wrote:

[...] This was my best shot at explaining what the alignment object is, with some explanation of how it got to be that way and why: http://blogs.pjjk.net/phil/explaining-the-lrmi-alignment-object/

Ah, I read Phil's blog and even commented on it shortly after it was written! So I can claim only forgetfulness, not ignorance ;)
 

Broadly speaking there are two reasons behind the AlignmentObject
(1) an extra level of indirection, which is useful when you don't want to specify every possible type of alignment in the properties of the thing that is aligned from or when you want to reify the alignment relationship in order to say some things about it (such as who says this resource is good for teaching that competency); and
(2) a surrogate for the thing that is being aligned to,  which is useful when there is no authoritative machine readable representation of the thing being aligned to.

And I guess we're mostly holding in our minds (in tension) *both* the fact that there are use cases which make AlignmentObject useful, *and* the fact that people seem easily confused by the double abstraction of EducationalAlignment together with AlignmentObject.

To me, it would seem that the ability to make either one of these abstractions concrete, rather than abstract, would make life easier for people trying to support or use LRMI in these cases.

1. As a parallel alternative (*not* a replacement) one could perhaps allow people to use a very small number of the most useful types of educationalAlignment, essentially taking recommended values of the alignmentType from the AlignmentObject. Sure, this would make little or no difference to the logic, just to the ease of understanding and use. Maybe it would be easier for people to see that something is a learning resource in virtue of having a straightforward "teaches" property, instead of an educationalAlignment to an AlignmentObject where the alignmentType is "teaches".

+1 This would help with the initial confusion caused by the abstractness of the indirection. There are a likely a few properties (teaches, assesses) that would cover a considerable percentage of cases. When used, the alignmentType property in the AlignmentObject need not be used.

When the nature of the relationship is not explicitly there as a property, then the general educationalAlignment property could be used to point to an AlignmentObject instance and the alignmentType property of the AlignmentObject could identify the nature of that relationship. It could also mean that defined properties like "teaches" and "assesses" could point directly to a framework node if there was no utility in a given instance of "reifying" the property (i.e., no desire to assert things like who asserts, strength of the assertion etc.) 

 

On the other hand, maybe aligning to an AlignmentObject doesn't capture what is being looked for. AlignmentObject points to an (node in an) educational framework, not a Thing in general, and it looks like people could use something that relates to a Thing.

Perhaps.

 

2. If we had some of these concretised educationalAlignment properties, we could allow suitable ones to have Things as their ranges, rather than AlignmentObjects? Sure, at present it wouldn't seem to make any sense to have an educationalAlignment directly to a Thing, but maybe that's because the type of alignment has been packed away in the AlignmentObject, instead of being an integral part of the educationalAlignment itself. And my understanding is that, well, it couldn't be...

Really hoping that this is more helpful than counterproductive, and with apologies in advance if people find the latter.

Simon, the discussion is not counterproductive. The issues have come up in the LRMI Task Group. Phil has raised them. I have raised them. People applying schema.org to learning resources have raised the difficulties they encounter in using the educationalAlignment property and the AlignmentObject type. 

Simon

--

from Simon Grant +44 7710031657 http://www.simongrant.org/home.html

--

[Renato]

Interesting discussion.

Simon, the discussion is not counterproductive. The issues have come up in the LRMI Task Group. Phil has raised them. I have raised them. People applying schema.org to learning resources have raised the difficulties they encounter in using the educationalAlignment property and the AlignmentObject type. 

 

Indeed, the (potentially powerful) property educationalAlignment seems to be rarely used (and quite "creatively").  For example:

                  addition+more:pagemap:creativework-learningResourceType produces 13900 hits

while

                  addition+more:pagemap:creativework-educationalAlignment produces just 7 hits.

The property learningResourceType seems to be much more useful to practically identify (by “inference” – the only possibility at this stage) a learning resources.

Renato

[Phil Barker]

On 25/01/2016 21:17, Renato wrote:

The property learningResourceType seems to be much more useful to practically identify (by “inference” – the only possibility at this stage) a learning resources.

Yes, I agree with this conclusion and the analysis that leads to it.