Raising the issue of collections, especially of learning resources, within schema.org

103 views
Skip to first unread message

Charles Myers

unread,
Sep 6, 2013, 12:08:16 AM9/6/13
to lr...@googlegroups.com
The Accessibility metadata thread on the W3C public vocabs list has heated up lately.  You can see this at http://lists.w3.org/Archives/Public/public-vocabs/2013Sep/ although my responses are currently blocked at the W3C, as I just started on the lists. You can also see the thread on the accessibility metadata google group at http://www.a11ymetadata.org/discuss/

While I was writing one of my emails on a hot thread today, I was reminded of a longstanding schema.org issue that I wanted to bring up to the LRMI group. Every LRMI workshop I've been to has a box of books as a package. See http://www.mcgraw-hill.co.uk/sra/rl_index.html for a good example. But the lament has always been that there is no such mechanism in schema.org for having a collection.

Well, that's wrong.  There is one.  It's http://schema.org/CollectionPage , but is expected to be a collection of web pages and exclsuive to that.  In fact, the way that it works is that WebPage has a property isPartOf, which then points to the collection page.

I believe that LRMI should consider bringing the concept of the collection and abstracting this to be a collection of any creative works, whether book, web page, app or video, to schema.org.  It's not a specific LRMI issue, but it comes up commonly in LRMI content.  

Also, do take a look in at the debate going on the public vocabs list. This is the first significant dialog we've had with the entire schema.org group, and your insights can help us get the best quality specification adopted, and then work on getting it applied to learning resources.

Phil Barker

unread,
Sep 6, 2013, 4:49:40 AM9/6/13
to lr...@googlegroups.com, Richard...@oclc.org

Hello Charles, hello all,
I mostly agree with subject line: we should be raising the issue of collections, it is not something special to learning resources, but perhaps we have special cases relevant to education that should be raised.

There was a proposal from schema bibliography group http://www.w3.org/community/schemabibex/ for collections http://www.w3.org/wiki/WebSchemas/Collection discussed on the public schemas list over the summer -- the latter part of the discussion is toward the end of this page http://lists.w3.org/Archives/Public/public-vocabs/2013Jul/thread.html

cc-ing Richard Wallis who may know where that proposal stands?

Furthermore....
I think there is a similar issue around events that comes up when we discuss courses. A lecture course is a series of educational events and it would be nice to be able to say so, but (I think) schema.org only knows about TV series. Put the two together and a "course" in a more general educational sense is a series or structured collection of educational events and/or creative works. If an educational course is a type of collection then collection has a wider domain than creative works.

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/groups/opt_out.


-- 
<http://www.icbl.hw.ac.uk/~philb/>



Sunday Times Scottish University of the Year 2011-2013
Top in the UK for student experience
Fourth university in the UK and top in Scotland (National Student Survey 2012)


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.

Steve Midgley

unread,
Sep 9, 2013, 5:22:38 PM9/9/13
to lr...@googlegroups.com, Richard...@oclc.org
Great topic. 

If we take on collections, I feel like I'd want them to let me express "ordered relations" similar to what Jason was talking about with the JSON-LD question recently.

For example, not only do I want a simple collection of LRMI resources, but I'd like to be able to express the collection in some type of order. So I can have all the units of instruction in a specified order (each unit might be an LRMI resource).. 

But it opens up a rabbit hole of complexity (does it support multiple branches of collections? what about named branches - one for "advanced" one for "more explanation"? Does it support nested collections, so "Course" has a bunch of units, each unit is a collection of resources, etc?)

Steve

Charles Myers

unread,
Sep 9, 2013, 6:25:00 PM9/9/13
to <lrmi@googlegroups.com>, <Richard.Wallis@oclc.org>
Steve,

  This is an even better reason for needed to have the collection be specified by "hasPart," as this can be an ordered set.  The child-parent relationship, isPartOf, is inherently non-ordered.

Also, I'd think that there's noting standing in the way of collections nesting.  However, I would not see them nesting in one massive HTML/XML/RDF/Microdata file, but that there would be accompanying href's to the next level down (so that the next level could, in fact, be a part of multiple collections).  Your branching is a whole other question, though.


On Sep 9, 2013, at 2:22 PM, Steve Midgley <steve....@mixrun.com>
 wrote:

Steve Midgley

unread,
Sep 9, 2013, 7:09:54 PM9/9/13
to lr...@googlegroups.com, <Richard.Wallis@oclc.org>
Thanks. 

If I can specify named, ordered relationships, then I can point from one collection to another with a name like "Advanced" which would help a lot..

Steve

Steve Midgley

unread,
Sep 9, 2013, 8:59:42 PM9/9/13
to Wallis,Richard, lr...@googlegroups.com, public-sc...@w3.org
Given my relative lack of expertise on the format side, I'll re-iterate some use-cases first and hopefully someone will have suggestions:
  1. Create a table of contents to a book, where each entry in the ToC is itself either another OrderedCollection or a URL to something more tangible (like a PDF or html page)
  2. Create a learning progression description (could be also a syllabus or curriculum)
    1. Each link in the OrderedCollection needs to describe the link itself in various ways
      1. Link metadata might be arbitrary but would probably include: link type (relationship) and required attributes
In a naive implementation, I'd guess I'd want:
  • A schema structure for links themselves
  • Extend ItemLists to LinkLists (to let me list Links in arbitrary order - in html probably ol/ul; in json probably arrays of hashes I guess)
  • Decorate the OrderedCollectionPage with LinkList objects

DataType(?!) > Link
[[all from CreativeWork?]]
relationship: text describing the link relationship
required: true if link is mandatory in context of completing a pathway in an OrderedCollection

Thing > CreativeWork > ItemList > LinkList
[Extends ItemList to permit Link object for each item in list]

Thing > CreativeWork > WebPage > CollectionPage > OrderedCollectionPage
linksTo: LinkList

Is that crazy?
Steve



On Mon, Sep 9, 2013 at 3:22 PM, Wallis,Richard <Richard...@oclc.org> wrote:
Copying Schemabibex list for information.

Steve,

My reaction is that an OrderedCollection should be a sub-type of Collection.

The question then comes to how you represent the ordering.   RDF gives us ref:first, rdf:rest, ref:nil, but I can't see how these would easily map into the schema.org world.

We could have first & last as properties for OrderedCollection, the challenge would be ordering the ones in-between.  

Any suggestions?


~Richard

Jason Hoekstra

unread,
Sep 9, 2013, 9:52:23 PM9/9/13
to lr...@googlegroups.com, <Richard.Wallis@oclc.org>
Hi all -

As Steve mentioned, I raised a related question in determining how to serialize LRMI in JSON-LD for use in systems like the Learning Registry.  If JSON-LD could be used for the work you are doing, the specification has a definition for sets and lists (6.11 Sets and Lists).  In that definition, lists are ordered collections.

In the other thread, Gregg Kellogg is helping me understand the @graph construct for linking data as well as the right attributes to define the relationships.  In meshing together the two concepts, this may be possible to get ordered relationships.  Borrowing from the example in the other thread and adding the @list keyword, the below would assert that "Solar System Exploration" is the primary item which has sub-items of item #1: "3-2-1 Liftoff! #01 Station Information" then item #2: "3-2-1 Liftoff! #02 Destination: Station".

{
    "@context": [
        "http://schema.org/",
        "dc": "http://purl.org/dc/terms/"
  ],
"@graph":
{
   "@list":
  [
    {
      "@type": "CreativeWork",
      "name": "3-2-1 Liftoff! #01 Station Information",
    },
    {
      "@type": "CreativeWork",
      "name": "3-2-1 Liftoff! #02 Destination: Station",
    }
  ]
},
    "@type": [ "CreativeWork" ],
    "name": "Solar System Exploration",
}

In regards to branching, that isn't possible as I understand the spec given this statement:  "List of lists are not allowed in this version of JSON-LD. This decision was made due to the extreme amount of added complexity when processing lists of lists.".  However, if each item is published individually, each could relate using "hasPart" using first-order links, then the linked items below continue the same pattern to establish the tree or graph you'd be looking for, such as:

Solar System Exploration hasPart #1: 3-2-1 Liftoff! #01 Station Information and hasPart #2: 3-2-1 Liftoff! #02 Destination: Station

3-2-1 Liftoff! #01 Station Information hasPart #1:  Station Body and hasPart #2 Station Docks and hasPart #3 Solar Panels and hasPart #4 Review Section 

3-2-1 Liftoff! #02 Destination: Station hasPart #1:  Connecting to Station and hasPart #2 Disengaging from Station and hasPart #3 Review Section

I'm coming up to speed with this as I read more into JSON-LD, but in this work, realizing the value of the serialization format to accommodate needs such as this.  Those with more experience of course much appreciated in validating / guiding the right way to represent this.

Jason

Steve Midgley

unread,
Sep 9, 2013, 10:25:47 PM9/9/13
to lr...@googlegroups.com, <Richard.Wallis@oclc.org>
Thanks Jason. Would that JSON-LD have an equivalent expression in schema.org microdata?

Steve

Gregg Kellogg

unread,
Sep 9, 2013, 11:25:58 PM9/9/13
to lr...@googlegroups.com, lr...@googlegroups.com, <Richard.Wallis@oclc.org>
On Sep 9, 2013, at 7:25 PM, Steve Midgley <steve....@mixrun.com> wrote:

Thanks Jason. Would that JSON-LD have an equivalent expression in schema.org microdata?

That depends, JSON-LD is a superset of what microdata can express when interpreted as RDF, if just transforming microdata to JSON, the topology is a bit different; JSON-LD doesn't use a "properties" key.

Gregg

Phil Barker

unread,
Sep 10, 2013, 8:07:07 AM9/10/13
to lr...@googlegroups.com
Hello all,
Good to have some simple use cases, and a warning about modelling complex processes such as those with conditional steps. Interestingly the same issues have been discussed over the last week on the main public-vocabs list in relation to an inquiry about checklists[1].

Of course these use cases aren't new, and so there is already work from which we might draw inspiration.  Two of these spring to my mind.

*  OAI-ORE (Object reuse and exchange) [2] is a sematic model for describing aggregations of web resources which was serialized by extending Atom. Use cases included papers in a journal, and images, data, sections within a paper. Very similar use cases to a table of contents. Ordering was less important to them but was considered, and lead to some complexity (it came down to the truthfulness of saying X comes after Y, which is true only in some specific context, which is hard to say when all you have is a triple in a global graph). As far as I know ORE hasn't been widely implemented, but the model seems functional.

* IMS Simple Sequencing [3] "defines a method for representing the intended behavior of an authored learning experience" which includes that rabbit hole of complexity, e.g. supporting use cases such as "if the learner passes this test then proceed otherwise go to these remedial exercises". I think this was perhaps more widely implemented since the sequencing and navigation element of SCORM 2004 was derived from it. I like the note by Rustici pointing out the the "simple" refers to scope not ease of implementation [4]. Beyond Simple Sequencing lies IMS Learning Design...

Anyway, my feeling is that any LRMI/Schema work on collections and courses as collections should be focused on description and discovery. Runtime requirements, issues such as ordering, sequencing, and flow control are complicated enough to be left to other specifications (that includes the use ol in HTML for ordering, for example). There does need (I think) to be a little more structure than a simple list. It does need to be hierarchical: the collection of resource used in a lesson can be part of a collection of lessons that is a module, which can be part of a collection of modules that is a course or programme--all I'm saying is that you need to have collections of collections, I don't think it will be at all useful to try to get into nomenclature to distinguish between what is a module/course/programme. Also there may be need for relationships other than hasPart, I don't think a recommended text for a course has the same relationship to that course as one of the lessons.

To be explicit, two/three use cases in that:
- to be able to find an educational collection of learning resources (creative works and/or events and/or educational collections) which is structured in such a way as to facilitate learning a specified topic/skill/competence. [this definition applies equally to courses and textbooks]
- to be able to identify the parts of a course and the resources external to that course which are referenced by it.

Phil


1. thread "Semantically marking up a "checklist" or process" at http://lists.w3.org/Archives/Public/public-vocabs/2013Sep/thread.html
2. http://www.openarchives.org/ore/
3. http://www.imsglobal.org/simplesequencing/
4. http://scorm.com/scorm-explained/technical-scorm/sequencing/

Steve Midgley

unread,
Sep 10, 2013, 7:54:17 PM9/10/13
to lr...@googlegroups.com
Hi Phil,

Very helpful - in fact based on your input, I would back off my original request for ordering and instead just ask that I can name the relationship or link between the Collection and each of its members. So even if I'm not talking about a semantically ordered list, I would like to be able to describe the relationship between a collection and the things it contains.

So I'd like:

Collection
   http://abc (Advanced)
   http://xzy (Read Me Second)
   http://def (Read Me First)
   http://ghi (If you're confused)

So my collection could list things out of order (or not control ordering), but each element could have a description or type which indicates how it relates to the collection. Obviously it would be better if these relationships were pulled from a controlled vocab of possible relationship types, but even with a free form text field it would open the door for some useful modeling that could coalesce into structured vocabs later.

Unless I'm thinking of this all wrong, which is a persistent possibility..

Steve



Phil Barker

unread,
Sep 13, 2013, 4:55:48 AM9/13/13
to lr...@googlegroups.com

Hi Steve, everyone.
A limited number of typed relationships between resources and course/collection is a good idea, I'm still nervous about trying to specify the ordering with LRMI (or schema.org in general). I think it is difficult and outwith schema's core use case of resource discovery via search engines.

How about:

Collection
   http://abc (optional advanced material)
   http://xzy (core material)
   http://def (core material)
   http://ghi (optional for if you're confused)
   http://pqr (describes the course structure)

http://pqr could be the organisation section from an IMS Content Package manifest, the spine section from an EPUB3 package, an IMS Simple Sequence, and IMS Learning design, OAI-ORE or any number of other things

Incidentally, this weeks mega discussion on the public-vocabs list has been about check-lists, where ordering/sequencing/processes has also been raised.

Phil


On 11/09/2013 00:54, Steve Midgley wrote:
Hi Phil,

Very helpful - in fact based on your input, I would back off my original request for ordering and instead just ask that I can name the relationship or link between the Collection and each of its members. So even if I'm not talking about a semantically ordered list, I would like to be able to describe the relationship between a collection and the things it contains.

So I'd like:

Collection
   http://abc (Advanced)
   http://xzy (Read Me Second)
   http://def (Read Me First)
   http://ghi (If you're confused)

So my collection could list things out of order (or not control ordering), but each element could have a description or type which indicates how it relates to the collection. Obviously it would be better if these relationships were pulled from a controlled vocab of possible relationship types, but even with a free form text field it would open the door for some useful modeling that could coalesce into structured vocabs later.

Unless I'm thinking of this all wrong, which is a persistent possibility..

Steve


Reply all
Reply to author
Forward
0 new messages