UMBEL v 0.80 Released

[mkbergman]

Hi All,

After some delays, we're pleased to announce the release of the new
UMBEL v 0.80 [1] and a new Web site [2] to accompany it. We look
forward to your testing and comments!

This is also the first result of our new partnership with Ontotext
[3]. With their expertise and resources, we should see a much
invigorated UMBEL effort moving forward. Thanks, Ontotext!

Please note this is but the first step leading to UMBEL v 1.00, slated
to be released by the end of the year. That version will include
better categorizatioin and mapping to Wikipedia (en) and PROTON [4],
and will also likely lead to some additional UMBEL language
refinements.

Thanks!

Mike

[1] a summary of major changes is provided on my blog announcement, at
http://www.mkbergman.com/930/announcing-a-major-new-umbel-release/
[2] http://umbel.org
[3] http://ontotext.com
[3] http://proton.semanticweb.org/

[Aleksander Pohl]

> > After some delays, we're pleased to announce the release of the new
> > UMBEL v 0.80 [1] and a new Web site [2] to accompany it. We look
> > forward to your testing and comments!
> >
> > This is also the first result of our new partnership with Ontotext
> > [3]. With their expertise and resources, we should see a much
> > invigorated UMBEL effort moving forward. Thanks, Ontotext!
> >
> > Please note this is but the first step leading to UMBEL v 1.00, slated
> > to be released by the end of the year. That version will include
> > better categorizatioin and mapping to Wikipedia (en) and PROTON [4],
> > and will also likely lead to some additional UMBEL language
> > refinements

I am very happy with this announcement.
And I am especially happy, with the fact, that you map other ontologies
to UMBEL (like FOAF and MO). Still I am curious, whether you are going
to map only concepts, or relations as well? I think, that only with the
latter, the true knowledge sharing will be possible. Also I am very much
interested in the methodology of mapping. I've read on the Google Code
page, that you use Cytoscape, but the procedure for finding
corresponding concepts is not described in detail.

Kind regards,
Aleksander

[mkbergman]

Hi Aleksander,

On Nov 19, 11:49 am, Aleksander Pohl <apoh...@gmail.com> wrote:
> > > After some delays, we're pleased to announce the release of the new
> > > UMBEL v 0.80 [1] and a new Web site [2] to accompany it. We look
> > > forward to your testing and comments!
>
> > > This is also the first result of our new partnership with Ontotext
> > > [3]. With their expertise and resources, we should see a much
> > > invigorated UMBEL effort moving forward. Thanks, Ontotext!
>
> > > Please note this is but the first step leading to UMBEL v 1.00, slated
> > > to be released by the end of the year. That version will include
> > > better categorizatioin and mapping to Wikipedia (en) and PROTON [4],
> > > and will also likely lead to some additional UMBEL language
> > > refinements
>
> I am very happy with this announcement.
> And I am especially happy, with the fact, that you map other ontologies
> to UMBEL (like FOAF and MO). Still I am curious, whether you are going
> to map only concepts, or relations as well? I think, that only with the
> latter, the true knowledge sharing will be possible.

As a reference ontology, we keep the internal relations as simple as
possible. It is possible for outside ontologies to relate their
properties as attribute mappings via the hasCharacteristic predicate,
which provides a means to access those relations, but that is not
actually modeling the total world view.

Again, we are not intending UMBEL to be a knowledge base, but a
reference ontology. It is merely a third-party point for saying some
content is "about something", really nothing more. "True knowledge
sharing" is much more ambitious than our intent to help get related
content into the same room.

Also I am very much
> interested in the methodology of mapping. I've read on the Google Code
> page, that you use Cytoscape, but the procedure for finding
> corresponding concepts is not described in detail.

We do not map using Cytoscape. We are using a combination of some
automatic techniques using things like the Alignment API and Semantic
Vectors, with much, much hand inspection.

Cytoscape is used to provide a rendering of the large graphs.

We will be documenting more on our mapping approaches after release of
UMBEL 1.0. For now, you can look to this Annex [1] as a beginning
point.

Thanks, Mike

[1] http://umbel.org/specifications/annex-f
>
> Kind regards,
> Aleksander

[Frederick Giasson]

Hi Alex!

> And I am especially happy, with the fact, that you map other ontologies
> to UMBEL (like FOAF and MO). Still I am curious, whether you are going
> to map only concepts, or relations as well? I think, that only with the

One thing to add here, in version 1.0 there will be a mapping to
Proton[1], and this kind of relationships will be available via this
Proton/UMBEL mapping. It is what they are doing with it to make queries
to FactForge [2] easier to formulate (with a cannonical ontology).

[1] http://factforge.net/
[2] http://proton.semanticweb.org/

Take care,

Fred

[Aleksander Pohl]

> As a reference ontology, we keep the internal relations as simple as
> possible. It is possible for outside ontologies to relate their
> properties as attribute mappings via the hasCharacteristic predicate,
> which provides a means to access those relations, but that is not
> actually modeling the total world view.

Ok, thanks for the clarification. As UMBEL is based on Cyc, I thought
a large portion of the relations is also available in UMBEL.

> Again, we are not intending UMBEL to be a knowledge base, but a
> reference ontology. It is merely a third-party point for saying some
> content is "about something", really nothing more. "True knowledge
> sharing" is much more ambitious than our intent to help get related
> content into the same room.

Yes. I understand that UMBEL is not a KB. But as an ontology, besides
classes it could have had more specific relations (like say homepage).
The "true knowledge sharing" for me is a scenario, in which I can
build an application based on some ontology (say UMBEL or Cyc) and
interpret the facts available in linked datasets, via mapping of the
concepts and the relations.

So far it seems for me, that the number of relations is smaller or
similar to the number of concepts (e.g. latest DC - 50 relations, MO -
120, DBpedia - 1200 relations, and only Cyc has ~20 000 relations) and
since the domain and range for theme should be clearly defined, having
the concepts mapped, it should be relatively easy, to find the
corresponding relations.
Anyway - now I understand, that this is not your design goal.

> We do not map using Cytoscape. We are using a combination of some
> automatic techniques using things like the Alignment API and Semantic
> Vectors, with much, much hand inspection.
>
> Cytoscape is used to provide a rendering of the large graphs.
>
> We will be documenting more on our mapping approaches after release of
> UMBEL 1.0.  For now, you can look to this Annex [1] as a beginning
> point.

Ok, thanks for the clarification and the pointer - I'll have a look at that.

Kind regards,
Aleksander

[Frederick Giasson]

Hi Alex,

> Yes. I understand that UMBEL is not a KB. But as an ontology, besides
> classes it could have had more specific relations (like say homepage).
> The "true knowledge sharing" for me is a scenario, in which I can
> build an application based on some ontology (say UMBEL or Cyc) and
> interpret the facts available in linked datasets, via mapping of the
> concepts and the relations.
>
> So far it seems for me, that the number of relations is smaller or
> similar to the number of concepts (e.g. latest DC - 50 relations, MO -
> 120, DBpedia - 1200 relations, and only Cyc has ~20 000 relations) and
> since the domain and range for theme should be clearly defined, having
> the concepts mapped, it should be relatively easy, to find the
> corresponding relations.
> Anyway - now I understand, that this is not your design goal.

Sure that you can re-use the OpenCyc relations as you want. Since we
have a 100% mapping with the core UMBEL, there is no issue re-using
them: the same semantic should apply.

What is currently demonstrated with these umbel linkages is that just
virtue of linking these external ontologies concepts to UMBEL, you can
use that coherent structure to make coherent assertions between
individuals instantiated using UMBEL (reference concepts) classes and by
re-using properties defined in external ontologies. By making these
assertions, we are really just putting in place the framework that will
help you to make sure that you are properly creating statements,
coherently according to the umbel conceptual structure.

On the other hand, what you seem to want is what will be available via
Proton: cannonical predicates that links to a set of related
relationships from different ontologies. This is another need which will
be available, via Proton, once the linkage become available in umbel 1.0.

Hope this helps clarifying this matter.

Thanks

Fred

[Aleksander Pohl]

Hi Frederick,

2010/11/20 Frederick Giasson <fr...@fgiasson.com>:

> On the other hand, what you seem to want is what will be available via
> Proton: cannonical predicates that links to a set of related relationships
> from different ontologies. This is another need which will be available, via
> Proton, once the linkage become available in umbel 1.0.
>
> Hope this helps clarifying this matter.

At least partially :-) I am not sure what is the relationship between
UMBEL and Proton. I guess there will be mappings between Proton
concepts and UMBEL concepts (in haven't found them so far in the
Appendix A). But UMBEL won't incorporate Proton relations. So it seems
for me, that Proton could become the canonical ontology I have in
mind. Or the taxonomy part will be available in UMBEL, and the
predicates (with their mappings) in Proton. Is that right?

Kind regards,
Alex

[Frederick Giasson]

Hi Alex,

this helps clarifying this matter.

> At least partially :-) I am not sure what is the relationship between
> UMBEL and Proton. I guess there will be mappings between Proton
> concepts and UMBEL concepts (in haven't found them so far in the
> Appendix A). But UMBEL won't incorporate Proton relations. So it seems

Exact, this mapping will be released for version 1.0 (before end of this
year) [1]

> for me, that Proton could become the canonical ontology I have in
> mind. Or the taxonomy part will be available in UMBEL, and the
> predicates (with their mappings) in Proton. Is that right?

Exact. The goals of these two ontologies are different, but the outcomes
are complementary. So, it is why we teamed up with OntoText for that
development (they are the creators and maintainers of Proton).

[1] http://umbel.org/nextchanges

Thanks,

Fred

[Aleksander Pohl]

Hi Frederick,

Thanks!

Regarding the "Mapping to Wikipedia Categories" task, I am not sure if
you are aware of the work done by Samuel Sarjant, Catherina Legg et
al., which is described in the paper "“All You Can Eat”
Ontology-Building: Feeding Wikipedia to Cyc". I am also not sure, how
much you have done so far with respect to the mapping, but I would
recommend looking into this paper.

Kind regards,
Aleksander

[Mike Bergman]

Hi Aleksander,

On 11/20/2010 12:50 PM, Aleksander Pohl wrote:
> Hi Frederick,
>
>>> At least partially :-) I am not sure what is the relationship between
>>> UMBEL and Proton. I guess there will be mappings between Proton
>>> concepts and UMBEL concepts (in haven't found them so far in the
>>> Appendix A). But UMBEL won't incorporate Proton relations. So it seems
>> Exact, this mapping will be released for version 1.0 (before end of this
>> year) [1]
>>
>>> for me, that Proton could become the canonical ontology I have in
>>> mind. Or the taxonomy part will be available in UMBEL, and the
>>> predicates (with their mappings) in Proton. Is that right?
>> Exact. The goals of these two ontologies are different, but the outcomes are
>> complementary. So, it is why we teamed up with OntoText for that development
>> (they are the creators and maintainers of Proton).
>>
>> [1] http://umbel.org/nextchanges
>>
> Thanks!
>
> Regarding the "Mapping to Wikipedia Categories" task, I am not sure if
> you are aware of the work done by Samuel Sarjant, Catherina Legg et

> al., which is described in the paper "�All You Can Eat�

> Ontology-Building: Feeding Wikipedia to Cyc". I am also not sure, how
> much you have done so far with respect to the mapping, but I would
> recommend looking into this paper.

Thanks; yes, we are quite familiar with this paper.

Mike
> Kind regards,
> Aleksander
>

--
__________________________________________

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