Adding tags from another ontology
Hugh Paterson III
I have a question about enriching resources with another ontology.
I am dealing with course content which is linguistic in nature and/or is in or about minority languages (Marked with ISO 639-3).
Linguistics has an ontology called ProjectGOLD [http://linguistics-ontology.org/] and we have implemented ISO 639-3 as an ontology as well.
It would seem that some of these Project GOLD Items could be used as subject tags, but they should also be marked as equaling terms in another ontology. How is that done? is that system dependent or does LRMI make provision for this?
I was also watching the LRMI tagger app video [http://www.youtube.com/watch?v=Sleh-ZGQ9mM] when I was think about this. We have a corporate repository (DSpace) and have an app kinda like tagger to implement/extract from submitters our metadata schema upon ingest.
But as we consider that LRMI metadata is being presented to search engines. It is generally presented in the HTML text like micro-data, Right? So, I am still wondering if and how these search engines will be able to make sense of terms (derive meaning) from other ontologies as they associate with LRMI terms.
- Hugh Paterson
UX Consultant
SIL International
Greg Grossmeier
Is the GOLD ontology for subjects easily reused by others without fee?
If so, then it could be a good option for those who want to fill in the
subject field with a controlled vocabulary (as you suggest).
Much discussion on the Schema.org community list (which, to remind
everyone, LRMI fits within Schema.org, thus inherits all of the aspects
there of) has indicated that the use of Wikipedia as a controlled
vocabulary for certain things also works well (though, obviously, not
with the breadth and depth as some domain specific ontologies out
there).
Best,
Greg
<quote name="Hugh Paterson III" date="2012-10-13" time="22:22:39 -0700">
Greg Grossmeier
Education Technology & Policy Coordinator
twitter: @g_gerg / identi.ca: @greg / skype: greg.grossmeier
Hugh Paterson III
Gold is free to use as is ISO 639-3. no fees :-)
But my question is a bit more specific than the extent to which you answered.
My questions comes around to an RDF description of a resource. My understanding of ontologies as implemented under RDF is that one has to define which ontology a descriptor token belongs.
How do RDF readers (like google) disambiguate subject terms. That is if I have the subject term "semantics" how is the RDF reader going to know if I mean the Wikipedia definition, the Project Gold definition, or a generic LRMI subject definition? at some point these things get embeded into the presentation layer as microdata right?
- hugh
Greg Grossmeier
<quote name="Hugh Paterson III" date="2012-10-15" time="17:09:35 -0700">
> if I have the subject term "semantics" how is the RDF reader going to
> know if I mean the Wikipedia definition, the Project Gold definition,
> or a generic LRMI subject definition? at some point these things get
> embeded into the presentation layer as microdata right?
understand subject terms is by the use of canonical URLs/URIs which,
ideally, are dereferencable with embedded metadata such that a machine
could learn more about it. That's the ideal.
Begin taking pieces out of that and you get less ideal, but mostly
workable.
Greg
Greg Grossmeier
enumerations (ie: controlled vocabularies):
http://www.w3.org/wiki/WebSchemas/ExternalEnumerations
Best,
Greg
<quote name="Greg Grossmeier" date="2012-10-16" time="15:36:21 -0700">
Brandt Redd
Erika B
Greg Grossmeier
Please do step up! No requirement for who can participate as long as you
want to participate! :)
I had it on my list for tomorrow, but if you want to beat me to it,
please do :)
And of course, if you want to get any feedback, we're pretty nice on
list or you can send it directly to me.
All the best,
Greg
<quote name="Erika B" date="2012-10-18" time="07:15:26 -0700">
>
> I'm a librarian lurker, who's had coursework on LCC & MARC cataloging. I
> have a second advanced graduate degree in Education Psychology, and my
> around assessment and competency alignment with course development teams,
> since we consult on every new and revised University course. I'm also
>
> Would the EducationalAlignment sample be something a new participator could
> pull together, or were you looking for long-time veterans to volunteer? If
> you were fine with me taking a stab at it, I could model the sample after
> your Common Core work.
>
> Thanks and Hello!
>
> Erika Bennett
>
>
Brandt Redd
- name: How to Round Integers
- author: Thag Simmons
- publisher: Contoso Learning Resources Inc.
- mediaType: Video
- educationalAlignment
- alignmentType: teaches
- educationalFramework: CCSS
- targetDescription: Use place value understanding to round whole numbers to the nearest 10 or 100.
- targetUrl: http://corestandards.org/Math/Content/3/NBT/A/1
Joshua Marks
Brant,
Very clear. But a couple of notes:
Regarding educaitonalAlignment’s TargetName, in this case I would suggest using the alpha-numeric code and/or GUID provided buy the CCSSO/NGA Center in the name. In this case I suppose that is “CCSS.Math.Content.3.NTB.A.1” or something like that.
Greg,
This is where the proposed educationalAuthority as an organization entity -> http://schema.org/Organization is still needed in the spec as an attribute of the educaitonalAlignment or perhaps the educaitonalFramework. I had thought you added it. This would add CCSSO/NGA center to the alignment and together the TargetName/GUID is sufficient to match the aligned skill even if the TargetURL is not from the CCSSO but rat the equivalent URL form AB (http://www.academicbenchmarks.org/search/?standard_id=20962&topic_id=2225022) , ASN (http://asn.jesandco.org/resources/S1143480) or any other look-up reference hosted by some system for the Common Core.
Joshua Marks
CTO
Curriki: The Global Education and Learning Community
I welcome you to become a member of the Curriki community, to follow us on Twitter and to say hello on our blog, Facebook and LinkedIn communities.
Stuart Sutton
This is where the proposed educationalAuthority as an organization entity -> http://schema.org/Organization is still needed in the spec as an attribute of the educaitonalAlignment or perhaps the educaitonalFramework. I had thought you added it. This would add CCSSO/NGA center to the alignment and together the TargetName/GUID is sufficient to match the aligned skill even if the TargetURL is not from the CCSSO but rat the equivalent URL form AB (http://www.academicbenchmarks.org/search/?standard_id=20962&topic_id=2225022) , ASN (http://asn.jesandco.org/resources/S1143480) or any other look-up reference hosted by some system for the Common Core.
Joshua, I may be misinterpreting what you are saying, so correct me if I am. I disagree somewhat with how you have framed this. While I certainly agree that educational-authority-as-organization would be good and useful information to have available, I'd assert that the "authority" in both cases you note would be CCSSO. The "authority" for the content of a JES&Co. data representation of a CCSSO assertion is not JES&Co. In the ASN, this notion of organizational authority is what is call the "jurisdiction" or "jurisdictional authority" defined as "A legal, quasi-legal, organizational or institutional domain of the entity mandating the use of the achievement standard within a given context--e.g., California, Singapore, NCTM." CCSSO is the one who has jurisdictional authority--it is the one who authored the content of the node and exercises authority over that content. In this regard, I would say that organizations such as JES&Co., Academic Benchmarks and EdGate are data intermediaries that neither exercise nor assert "authority" with regard to the CCSSO-authored content of the data they mediate. I would even go so far as to assert that CCSSO is the final authority as to whether the JES&Co. representation is an accurate representation of its creation.
Now, if you want to know who the data mediator is (as opposed to promulgator :-) that is bringing that node of CCSSO data to you or your system, that's a different matter. But it is not a matter of "educational authority"...I would not think.
Stuart
Joshua Marks
Stuart,
I think we are in agreement, _somewhat_. The educationaAuthority for the framework should be CCSSO/NGA Center for the CCSS framework, regardless of the URL provider for that framework. Jurisdiction might be an more descriptive name then educationalAuthority. Or perhaps even more accurate for this use is something like frameworkAuthority or frameworkPublisher as an Organization. CCSSO does not really fit your definition of jurisdiction, does it? The adopters of the standard are the jurisdiction, which are the states and districts applying the CCSS. Hmmm.
Joshua Marks
CTO
Curriki: The Global Education and Learning Community
I welcome you to become a member of the Curriki community, to follow us on Twitter and to say hello on our blog, Facebook and LinkedIn communities.
From: lr...@googlegroups.com [mailto:lr...@googlegroups.com] On Behalf Of Stuart Sutton
Sent: Thursday, October 18, 2012 11:59 AM
To: lr...@googlegroups.com
Subject: Re: Adding tags from another ontology
On Thu, Oct 18, 2012 at 11:03 AM, Joshua Marks <jma...@curriki.org> wrote:
Jim Goodell
The collective effort of every publisher adding tags for framework, framework ownership, competency to comprtency relationships, etc. on every tagged resource is huge compared to the limited number of framework promulgators adopting a common markup, and adding a few tags to existing data driven pages that use info already in their database.
So I will reassert that it would be best to promote a standard markup and vocabulary for the aligned thing in an educatonFramework (tags within the targetUrl page). Joshua, I understand that this is beyond LRMI but the strategy has advantages for LRMI stakeholders, it lessens the burden on publishers, more consistent results for consumers, lets standards 'promulgators' do what they do.
...to Greg's point schema.org already has a way of dealing with external enumerations like LCC and language codes...and ownership, etc. so it may not be a matter of inventing any new terms, just a matter of establishing the convention of what to expect for the markup on the page that a targetUrl points to when it is to an item in an educational framework.
Erika B
In answer to Brandt’s question about other taxonomies….. LCSH was the most highly emphasized taxonomy, of course. This is a pretty good summary of our world of taxonomies: http://www.slideserve.com/anila/knowledge-organization-library-tools-and-taxonomies-for-the-web
I'm the subject specialist for our Business & Education programs, which is where classification can start to diverge. Discipline librarians focus on specialized information exchange systems like SIC, NAICS, etc. There are individual educational taxonomies from researchers like Patricia Alexandria, but you’ve probably already taken those into consideration.
The most frequent education-oriented taxonomy that I use is ERIC subject headings. But any subject heading classification can be notoriously out of date at times. I might be forgetting an obvious classification system, though. I'll noodle it over!
Another thing to consider is that library science cataloging coursework is changing all the time. I was in courses in 2004, which is practically eons ago, from a web cataloging standpoint. Here is an example of what’s being taught these days:
Mike Collett
But...
If re-use of identifiers of competences, objectives, topics etc between frameworks is to be encouraged then it is important to know the context that the identified concept comes from, and possibly its unique relationships in that context, not just its identity and resolution bundled together.
Competency X may be part of different qualifications in different frameworks or may be available translated for a different audience. In one framework it may be deprecated and in another still very topical.
It can also be useful to know the "full set" of possible terms when building interfaces for facetted classifications. Tagging something as Algebra from the set {algebra, number, geometry} does not have the same connotation as tagging it from a more granular taxonomy of 1000 mathematical terms.
Providing, the option at least, of being able to identify a concept *and* the set that it comes from is a good idea.
Cheers
Mike 7:-D
-----------
Mike Collett, Schemeta
+44 7798 728 747
------------
www.schemeta.com
email: mi...@schemeta.com
twitter: @schemeta
skype: mikecollett
people are the network
Jim Goodell
...we use terms that describe more complex relationships (ASN has adopted such a vocabulary) such as exact match vs part of vs 'same concept at a lower developmental level', that may go beyond your concern, but we can say something about threlationship between e.g. "Use place value understanding to round whole numbers to the nearest 10." Compared to "Use place value understanding to round whole numbers to the nearest 10 or 100."
Stuart Sutton
Should the framework, authority and other important attributes all be in the alignment object or should it be discoverable from the targetUrl? It is much easier for publishers of learning resources to tag content with targetUrl and trust ASN, AB, corestandads sites to have the other important information such as ownership/publisher/authority of the framework and relationship of the learning standard item to others in the framework, e.g. "This competency is a child of ...targetUrl". ...and it avoids data quality issues such as one tagger saying the related competency is part of the "ccss" framework and another saying it is part of "common core state standards for mathematics"
Jim, I agree and believe that LRMI provides the means for getting there. Reading the tea leaves, the attempt with LRMI so far has been to provide a minimal set properties with educationalAlignment to handle descriptive "strings" while providing through the targetURL some means for handling "things". The core data infrastructure to support an ultimate (hopefully inevitable) migration from "strings" to "things" with the various frameworks of import to educationalAlignment is rapidly developing.
Your scenario makes several useful and quite logical assumptions: (1) that the applications consuming the LRMI metadata (here the search engines) are interested in, and willing to, dereference the targetURL, (2) that they are interested in doing something useful with the any data (if any) that's returned; and (3) that what is returned is in fact useful data. There is currently some level of uncertainty around all three assumptions.
If we dereference the targetURL in Brandt's example above--i.e., the CCSS URI http://corestandards.org/Math/Content/3/NBT/A/1, we should get useful data. What we get back in this case is a web page that contains enough information to support the values we see in Brandt's educationalFramework and targetDescription properties. There is not much else we can know directly from the page about this alignment that Brandt hasn't already told us in the LRMI metadata without our going on a fishing expedition from this webpage. So, deferencing this URI certainly doesn't give us the kinds of contextual information you describe beyond providing a means of verification. (However, see an XML version in the pipeline: http://www.corestandards.org/Math/Content/3/NBT/A/1.xml).
It is quite possible for that targetURL to serve the purposes you suggest--i.e., to feed back "other important attributes" so they need not be replicated as strings in the LRMI metadata. As I said above, I think that LRMI is working to hit a middle ground so the metadata is sufficiently useful without the presence of the targetURL or without dereferencing one while nevertheless providing the data means for playing in the world of things. From the beginning, LRMI has struggle, and rightly so, with what that "useful middle ground" looks like.
But building the data infrastructure to support next steps in the education/training domain is already happening. For example, here's the same CCSS node in the ASN:
Simple view of browser deference (without ASN resolution services or SPARQL endpoint):
- -http://asn.jesandco.org/resources/S1143480
Notice the mapping to CCSSO identifiers in the data!
Data views:
-- As RDF/XML (URI target node + CCSS Math doc description)
http://asn.jesandco.org/resources/S1143480.rdf
-- As RDF/XML (URI target node positioned in taxon path (node's tree) + CCSS Math doc description)
http://asn.jesandco.org/resources/S1143480_taxon.rdf
Stuart
--
Stuart A. Sutton,
CEO and Managing Director, Dublin Core Metadata Initiative
Associate Professor Emeritus, The Information School
University of Washington
Brandt Redd
Joshua Marks
Brandt,
I certainly share your concern. However that does not mitigate the fact that there are already multiple data look-up representations for the common core that exist, and are being used by different groups and orgs. To complicate this further, granular decomposition of the lowest level skills in the common core seems to be needed and re-sequencing and organization is in play as well at the state level and within the assessment consortia. But you would have better visibility to this then I. This is why the GIM-CCSS initiative is so important, to avoid the potential fragmentation of the core. None the less, I see no way around the fact that there will be multiple data sources for the CommonCore standard used by system vendors and publishers right now. They ALL should provide and reference the canonical (unique) URI from the CCSSO or some GIM-CCSS related service if one is defined. How this will express the further granular decomposition and alignment is completely unknown at this time and the goal of the GIM-CCSS initiative as I see it.
Joshua Marks
CTO
Curriki: The Global Education and Learning Community
I welcome you to become a member of the Curriki community, to follow us on Twitter and to say hello on our blog, Facebook and LinkedIn communities.
Brandt Redd
Joshua Marks
I think the inter-framework relationship stuff is what Jim is pointing to.
Joshua Marks
CTO
Curriki: The Global Education and Learning Community
US 831-685-3511
Stuart Sutton
Good point! I agree with your description of the problem.I've been thinking about how to use the Learning Registry for this sort of thing. You could post records that say something like "ABC is equivalent to XYZ" where ABC and XYZ are URLs. Other relationships might be "is a superset of" and "is a subset of". The latter two relationships would be valuable for the GIM-CCSS project. By using the Learning Registry, the data would be available to all vendors -- either to support automatic retagging or on-the-fly translation.Of course, someone would have to define the schema for these records as the Learning Registry only manages the transport. I believe that the ASN folks are planning to do something like this.
Yes, Brandt, that is correct. Back in 2006-2007, ASN worked with the NSDL community under NSF funding that included development of initial mapping predicates based in emerging practice These mapping predicates have evolved and define relationships between competencies, or learning outcome statements (LOS) in separate competency sets. That work has now accelerated given today's stronger interest--first in machine-readable competency sets and second in building useful mappings on top of them. We see such mappings in the AAAS/NSDL strand maps [1] and in learning trajectories in The Netherlands and elsewhere. The 9th EdReNe Strategic Seminar at The Hague in early December will be looking at some of the issues and possibilities [2].
Some mapping predicates from this earlier work have been integrated into the ASN RDF schema at http://purl.org/ASN/schema/core/ (see sections in the schema dealing with mapping) and are being integrated into ASN mapping tools. IMHO, the development of useful mapping predicates will be an evolving proposition as people working with machine representations of competency sets such as the CCSS (among many, many, many others) find utility in mapping across LOS silos--utility in terms of downstream big data benefits as these mappings increasingly support adaptive learning environments and data intensive individual student learning plans.
Stuart
[1] http://strandmaps.nsdl.org/
[2] http://edrene.org/seminars/seminar9haag.html
Paul Libbrecht
Stuart Sutton
That said, I'll babble on a bit. The mapping predicates defined so far by ASN are certainly not exhaustive--thus my comment that this is an evolving space as more types of relationships are explored in practice and formally defined in the public sphere. Also, since it is a developing space, ASN documentation of semantics and use of the mapping predicates are also evolving. I pointed you to the RDF schema in my earlier email; but, a little fuller (slightly more human) picture is available in the ASN application profile documentation at [1]. However, I'd say that elimination of all ambiguity in this sort of mapping, while a worthwhile goal, is not likely to happen given the inherently subjective nature of mappings.
Additionally, all such mapping predicates do nothing more than assert certain kinds of relationships between two entities. In this sense, the level of nuance is limited although formal semantics can take a good distance. Where richer description is needed--i.e., fuller description of a specific relationship, then we're no longer talking about useful mapping predicates alone. The transition from identification to description is what led LRMI discussions from simple binary assertions like Learning Resource==teaches==>Competency to an alignmentObject where such a relationship can be more or less richly described.
Paul, I'm not sure how to respond to your question about "failure"--do you have particular kinds of failure in mind?
I reiterate, this thread has morphed into a conversation that should be happening elsewhere and in the context of other work...I think.
Stuart
[1] http://standards.jesandco.org/wiki/ASN_Application_Profile