noob questions

[Robert Forkel]

hi all,
i'm a programmer with the Max Planck Digital Library (the one behind
http://wals.info/) and just starting with GOLD, because i'd like to
make the WALS data - and that of another project - available as linked
data (http://linkeddata.org/). But trying to model the data using
GOLD, i ran into some problems.

1. in the "inLanguage" definition, the comment talks about relating
LinguisticSign and Language, while the domain property points to
LinguisticUnit.
2. i found the above problem, because i'm looking for the correct
concept for word, and am not sure whether to pick LinguisticSign or
just LinguisticUnit. I'd tend to LinguisticSign, related via hasForm
with a LinguisticUnit which in turn is inLanguage?
3. the range of hasForm is PhonologicalUnit, but this class is not
defined.

the new project i'm working on deals with Loanword Typology (http://
www.eva.mpg.de/lingua/files/lwt.html), so i'd also need to model loan
relationships. are there any extensions planned in this direction?

regards,
robert

[S. Farrar]

HI Robert

Thanks for your questions. I'm glad to know how folks are using the
ontology. Comments below:

>
> hi all,
> i'm a programmer with the Max Planck Digital Library (the one behind
> http://wals.info/) and just starting with GOLD, because i'd like to
> make the WALS data - and that of another project - available as linked
> data (http://linkeddata.org/). But trying to model the data using
> GOLD, i ran into some problems.

Are you linking up the WALS data from indiv. contributors?

>
> 1. in the "inLanguage" definition, the comment talks about relating
> LinguisticSign and Language, while the domain property points to
> LinguisticUnit.

right, I'll change this.

> 2. i found the above problem, because i'm looking for the correct
> concept for word, and am not sure whether to pick LinguisticSign or
> just LinguisticUnit. I'd tend to LinguisticSign, related via hasForm
> with a LinguisticUnit which in turn is inLanguage?

What was intended w LinguisticUnit is any kind of unit: 'form units' as in
a phonological form, ' semantic unit' as in a semantic category like Dog,
or grammatical unit (for lack of a better word) as in a syntacticword or
mopheme. So LinguisticSign might still be useful from an ontological
modeling standpoint. Practically speaking, LingUnit works better from the
implementation side.

> 3. the range of hasForm is PhonologicalUnit, but this class is not
> defined.
>
> the new project i'm working on deals with Loanword Typology (http://
> www.eva.mpg.de/lingua/files/lwt.html), so i'd also need to model loan
> relationships. are there any extensions planned in this direction?

possibly. We have a lexicon oriented project coming up and there could be
significant overlap here.

Scott

>
> regards,
> robert
> >
>

[Robert Forkel]

On Dec 4, 6:38 am, "S. Farrar" <far...@u.washington.edu> wrote:
>
> Are you linking up the WALS data from indiv. contributors?
>

no. just the core data. i.e. the data from the original authors. we
are thinking about ways to allow individuals to contribute to WALS,
but have not come to a conclusion so far.

[S. Farrar]

Hi Robert, Jeff, et al.

>
> no. just the core data. i.e. the data from the original authors. we
> are thinking about ways to allow individuals to contribute to WALS,
> but have not come to a conclusion so far.

If you do decide to use GOLD, then I'm developing software to make it
easier for non-programmers to migrate to a SemWeb framework. In
particular, I've built a Leipzig Gloss reader that will transform glossed
texts to RDF that links to GOLD. All a user has to do is put their data in
a text file in Leipzig style. My software is at e-linguistics.org and it's
in collaboration w Steve Moran who you might know. I started this during
my work w Jeff.

Now about data:

----------
Data structures
--------------

For lexical entries, we need to come up with agreed upon data structures
in OWL, like LexicalEntry, Lemma, etc. We'll be looking into this for one
of our other projects so it's definitely on the list.

Also, it's good you're using concepts like these for the actual content of
the data structure:

SyntacticWord, PhonologicalWord

That way, it's clear what kind of unit you're referring to. No matter what
kind of linguistic unit you have, you still have relations such as

orthographicForm, hasConstituent, etc

So, some word may be both an instance of SyntacticWord and
PhonologicalWord at the same time, like:

German 'Boot'

But we might have something that's only an instance of a PhonologicalWord,
like:

French 'l'ami'

This actually gets around needing a unifying concept like LinguisticSign.
Though in principle, one could have both.

So, what do you meaning by lwt:word ?

------
Meaning:
-------

My 2 cents on meaning. We need very general categories in GOLD, as general
as WordNets top categories, if you're familiar w that resource, things
like:

Animal, Plant, Action, LivingBeing, etc.

There are several possible starting points for this and I plan to include
these soon. So, for particular lang's we'd create COPEs (or
sub-ontologies). So if a language had an entry for 'cormorant', we'd
create a class in the COPE that inherited from gold:Bird, or some such.

----------
Languages/ varieties, languiods
------------

Just to stir the pot, please note that I've established a namespace for
Ethnologue entiries:

http://linguistics-ontology.org/languages/

And at this address you'll find an RDF file containing all of
Ethnologue15:

http://linguistics-ontology.org/languages/ethnologue15.rdf

---------
Other issues
------------

Jeff is correct. GOLD is very much a work in progress. We've been awaiting
developers like yourself to prod us into adding needed concepts. If you
decide to use it, then I can add what you need. Let's use this forum to do
so. In fact, I'm in the process of transferring the GOLD project to Google
code which will offer developers a more active role. More on this after we
make some admin decisions.

So, I'm in agreement w Robert who says that the app might drive the
development of the SemWeb resource. That's exactly the kind of interaction
we'd like.

more in a bit

Scott

[Robert Forkel]

>
> If you do decide to use GOLD, then I'm developing software to make it
> easier for non-programmers to migrate to a SemWeb framework. In
> particular, I've built a Leipzig Gloss reader that will transform glossed
> texts to RDF that links to GOLD. All a user has to do is put their data in
> a text file in Leipzig style. My software is at e-linguistics.org and it's
> in collaboration w Steve Moran who you might know. I started this during
> my work w Jeff.

nice to see you program in python, too. that should make things
easier. i'll definitely have a look at your package.

>
> So, what do you meaning by lwt:word ?

lwt:word is just the prefix of all URLs in LWT for words. it's the
subjects of the triples, not concepts.