Re: [ontolex] Modellin collocations

Hi Thierry,

The difference in modelling here is to do with what is causing the form to
change. In the case of 'estación de autobuses', the third word cannot be
affected by anything outside of the term and as such is always in plural.
In this case we model it using the decomp model as it is *inherent to the
term*.

In contrast we have forms for which the lexical entry can take, for example
the plural form or a particular case (accusative) which is imposed *externally
to the term*, that is from the context it appears in the sentence or a
particular meaning. A form restriction is a not that a particular entry
(term) cannot be used in a context that would force it to take a restricted
form, most typically that it cannot be used in the singular form.

Regards,
John

On Sat, 13 Jul 2019 at 10:54, Thierry Declerck <declerck@dfki.de> wrote:

> Dear John, Julia, All,
>
> @John: so yes, I can add the information to the component (stating that it
> is accusative and plural). But at the same time there is a kind a
> prescriptive information: "živali" has to be accusative and plural (similar
> to the example given by Julia for *estación de autobuses* (thanks Julia
> for extending the language scope :-)), as there is no hospital for just one
> animal, child, person etc. The form restriction is here not on the full MWE
> but only on the third component ("živali").
>
> Cheers
>
> Thierry
> Am 11.07.2019 um 18:16 schrieb John P. McCrae:
>
> Hi Thierry, Julia, all,
>
> I think this is what the decomp module does:
>
> https://www.w3.org/2016/05/ontolex/#components
>
> The example given there is the Spanish 'comunidad autonoma', where the
> adjective 'autonomo' has a feminine agreement. The modelling is done by
> adding LexInfo properties to the component.
>
> This is different from FormRestrictions, which also limit the meaning of a
> word to only applying to certain forms, e.g., 'goods' is only used in the
> plural. In "bolnišnica za živali", I assume the head (bolnišnica) is free
> to take any number or case and hence there is no lexico-semantic
> restriction on the multiword expression.
>
> Regards,
> John
>
> On Thu, 11 Jul 2019 at 17:02, Julia Bosque Gil <jbosque@fi.upm.es> wrote:
>
>> Hi Thierry,
>>
>> If I understood this right, in this case we are not dealing with
>> morpho-syntactic features of an entry when used in a specific
>> ontolex:LexicalSense (we would use lexicog:FormRestriction in that
>> case), but you just want to indicate that *živali* in *za živali* (in
>> the MW *bolnišnica za živali* ) must be in accusative plural. If that is
>> the case,  I am wondering whether here we could just indicate that the
>> second decomp:subterm of *bolnišnica za živali* is the lexical entry *za*,
>> and describe that preposition by indicating that it assigns accusative case
>> to its nominal complements. At first it looks like the fact that *živali*
>> is in accusative plural depends mostly on the use of *za* (?). If so,
>> the restriction on *case* we are referring to is not particular to this
>> MW,  but concerns the grammar of prepositions. For *bolnišnica za živali*
>> a potential way of encoding this information would be to list the
>> components in order and indicate at the :za lexical entry level that it
>> requires accusative (or turning to the synsem module to specify a
>> lexinfo:NounPPFrame to record that *bolnišnica *takes a PP headed by a
>> *:za* marker).
>>
>> I am struggling with the *plural* feature and where to record this,
>> though: in Spanish,* estación de autobuses* (bus station) requires
>> plural in *autobuses (?? estación de autobús)*, but *parada de autobús *(bus
>> stop) is used in both singular and plural, probably because of the
>> difference between a "station" (always for more than one bus line) and a
>> "stop" (for at least one bus line). So I agree there should be a mechanism
>> to represent that *bolnišnica *takes a za PP with a noun in plural,
>> maybe this would be information to address with synsem elements, as well?
>>
>>
>> What do you think?
>>
>> Best,
>>
>> Julia
>>
>> El jue., 11 jul. 2019 a las 14:47, Thierry Declerck (<declerck@dfki.de>)
>> escribió:
>>
>>> Dear All,
>>>
>>> I had just some sessions with Simon Krek on Slovenian Collocations/MWE
>>> and how to represent those in OntoLex-Lemon.
>>> I think for most of the data we have good solutions.
>>> Only a bit unclear to me how to model restrictions that are to be
>>> applied on a Collocation/MWE. An example is the expression "Hospital for
>>> Animals" (bolnišnica za živali {@slv}).
>>> On the third component of this expression there is a form restriction:
>>> it should be used only in accusative and plural.
>>> I would ot use the restriction mechanism described in Lexicog, but then
>>> the domain of the property would be a decomp:Component (and the range is
>>> a form, as foreseen)
>>> Another thing that could be useful: define the restrictions
>>> independently of a specific form. It would just list the two relevant
>>> aspect here: plural and accusative. Then we would need a mechamis that
>>> goes from the lemma to the corresponding form (the lemma, or headword,
>>> is given by the decomp:subterm property.
>>>
>>> Cheers
>>> Thierry
>>>
>>>
>>
>> --
>>
>> Julia Bosque Gil
>> PhD Student
>> Ontology Engineering Group <http://www.oeg-upm.net/>
>> Departamento de Inteligencia Artificial
>> Universidad Politécnica de Madrid
>>
> --
> Thierry Declerck
> Senior Consultant at DFKI GmbH, Multilinguality and Language Technology
> Stuhlsatzenhausweg, 3
> D-66123 Saarbruecken
> Phone: +49 681 / 857 75-53 58
> Fax: +49 681 / 857 75-53 38
> email: declerck@dfki.de
>
> -------------------------------------------------------------
> Deutsches Forschungszentrum für Künstliche Intelligenz GmbH
> Trippstadter Strasse 122, D-67663 Kaiserslautern, Germany
>
> Geschäftsführung:
> Prof. Dr. Jana Koehler (Vorsitzende)
> Dr. Walter Olthoff
>
> Vorsitzender des Aufsichtsrats:
> Prof. Dr. h.c. Hans A. Aukes
> Amtsgericht Kaiserslautern, HRB 2313
> -------------------------------------------------------------
>
>