Umbel And Linked Data

[Myriam Leggieri]

Hi everyone,
I read about Umbel in an article by Mike Bergman, understanding that
Umbel can be used to reduce the amount of links created in the context
of Linked Data. That's because one can link an Umbel concept and then
retrieve all the other datasets linking it instead of having a
centralized list of datasets on the same topic.
For example
if I have to link my data about bugs to all the other related bugs
stored in other external datasets available on the web and I neither
know which are these datasets nor want to keep a list of them to crawl
(through Sparql endpoint), then I could simply search for instances of
the Umbel subject concept 'sc:BugReport' (a umbel:SubjectConcept).

Have I understood well the application of Umbel to Linked Data?
I'm asking this question because in this mailing list and on the
Frederick Giasson's weblog I read about a completely different kind of
Umbel usage which consists in a support to the creation of new
ontologies, inserting them coherently in their proper context.

thanks in advance to anyone who would be so kind to answer me
cheers
myriam

[Myriam Leggieri]

maybe what I've understood is the same as what Frederick has written
in the Conclusion section of his blog post
http://fgiasson.com/blog/index.php/2008/09/04/exploding-dbpedias-domain-using-umbel/

[Myriam Leggieri]

ooo welll sorry for the spam XD
I've just found also this statement:
"Linked Data in context — fully exposed as Linked Data to provide a
conceptual graph for other Linked Data to relate and interconnect,"
within the featured page 'Introduction to Umbel'
http://groups.google.com/group/umbel-ontology/web/introduction-to-umbel
which suggests me that what I understood maybe is correct.

Then,
if I want to crawl through Sparql endpoint all the hosts publishing
linked bug data (supposing that all these hosts have linked themself
to the proper Umbel subject concept) I simply have to search for
instances (named entities) of the sc:BugReport class and try some
sparql queries over the hosts contained in the obtained resultset. Is
this right?
But then a list of all the sc:BugReport instances have to be stored on
the Umbel servers, and how could this happen? I should create a
statement in which I say that my ontology isAlignedWith (for example)
sc:BugReport but, how could this be enough? I should let Umbel know
about the existence of this statement... can I receive any clear
explanation about how to implement this particular kind of Umbel
usage?

On Jul 16, 3:46 am, Myriam Leggieri <myriam.leggi...@gmail.com> wrote:
> maybe what I've understood is the same as what Frederick has written

> in the Conclusion section of his blog posthttp://fgiasson.com/blog/index.php/2008/09/04/exploding-dbpedias-doma...

[park]

Please cancel my subscription. I can't seem to do so
through your Unsubscribe, below.

[Mike Bergman]

Hi Myriam,

Sorry for the late response.

If I understand your question correctly, I think there are two
parts to the answer: 1) what are the purposes of UMBEL; and 2)
how to use it.

UMBEL can indeed meet the purpose you outlined. Though other
usages might be listed, here are the four main uses that I refer
to for UMBEL:

1) providing a subject "concept" for characterizing the context
for content in different data sources -- your stated interest
2) for providing class linkages into external ontologies, what
you referred to from one of Fred's posts and he has called
"exploding the domain"
3) for inferencing, which allows expansion over subject concepts, or
4) for faceted searching.

UMBEL itself does not make the linkage assertions for related
content, as you supposed. Rather, UMBEL provides a vocabulary of
properties (predicates) for how to state contextual relationships
amongst instances and classes (concepts) and provides a
pre-defined list of 20 K subject concepts and their internal
relationships for reference.

So, for bug reports, others will need to have made the assertion
that their own information is aligned or equivalent to
sc:BugReport (in which case your SPARQL example holds) *or* you
will have discovered such similar content in which case you can
make the assertions locally.

If this is not a complete enough response, please let me know and
we can expand further.

Thanks!

Mike

--
__________________________________________

Michael K. Bergman
CEO Structured Dynamics LLC
319.621.5225
skype:michaelkbergman
http://structureddynamics.com
http://mkbergman.com
http://www.linkedin.com/in/mkbergman
__________________________________________

[myriam leggieri]

Hi Mike!
it's a honor for me to talk with you, compliments for your work :)
Then if I describe my dataset for example through a statement such as

dcterms:subject sc:BugReport

and make a sitemap using the sitemap extension "Semantic Web Crawling"  

containing a reference to my dataset description;
I could retrieve all the datasets who has the same subject as mine using a 
semantic search engine.

Is this the way in which Umbel could help me to find relate-able datasets?

thanks so much ;)
myriam

[Mike Bergman]

Hi Myriam,

Thanks for the kind words.

I have asked Fred to respond to you more in depth. Your Google
SoC project [1] looks quite interesting and a good potential use
of UMBEL.

Fred should respond shortly with a direct answer to your
questions and some other ideas. ;)

Thanks!

Mike

[1] http://code.google.com/p/hackystat-linked-sensor-data/

> <mailto:myriam.leggi...@gmail.com>> wrote:
> >> maybe what I've understood is the same as what Frederick has written
> >> in the Conclusion section of his blog
> posthttp://fgiasson.com/blog/index.php/2008/09/04/exploding-dbpedias-doma.

> <http://fgiasson.com/blog/index.php/2008/09/04/exploding-dbpedias-doma.>..

> >>
> >> On Jul 16, 2:06 am, Myriam Leggieri <myriam.leggi...@gmail.com

[myriam leggieri]

Hi!
I'm really glad that you've found my project interesting ^_^
I'm encountering some obstacles due to the lack of suitable external datasets to link but, at least, all the Hackystat data could be linked with each other :p

I'm excited to see what Fred is going to suggest me
deeply thanks to you both for your interest and help ^___^
myriam

[Frederick Giasson]

Hi Myriam!

> it's a honor for me to talk with you, compliments for your work :)
> Then if I describe my dataset for example through a statement such as
> |dcterms:subject |sc:BugReport
>
> and make a sitemap using the sitemap extension "Semantic Web Crawling"
>

What kind of sitemap? (there are a couple floating around; something you
will create?).

> containing a reference to my dataset description;
> I could retrieve all the datasets who has the same subject as mine using a
> semantic search engine.
>

Yes exactly, this is certainly something that is possible.

Also, if you have some basic inferencing enabled, you will be able to
query for the parent subject concepts of this concept and still getting
the data related to these tagged datasets.

It really depends on the way your system will work. You can do this kind
of "semantic tagging" of subject concepts at the level of the
description of your datasets, or you can go further an do this directly
with the types of the resources from each dataset.

> Is this the way in which Umbel could help me to find relate-able datasets?
>

Certainly. Once your datasets are properly tagged, you will be able to
leverage the entire UMBEL subject concept structure to enable some (at
least basic) inference. An example of this is if you query the structure
to get all the datasets related to "Projects", then you will get all the
datasets that have been "tagged" with more specific concepts such as
"construction project", "research project", "anthropological studies", etc.

Please, ask me any other questions you could have related to this since
this is quite important for you project and UMBEL.

Thanks,

Take care,

Fred