Re: morph module feedback on assigning meanings

Dear Bettina, dear all,

some comments below.
>
> Tuesday the issue of assigning meanings to morph resources has been  
> addressed within the telco. For the following two issues we discussed  
> >possibilities but could not reach a convincing conclusion. I kindly ask  
> you to provide your feedback on this.
>
> 1) Describing the meanings of stem resources
>
> Option one is to create a new property morp:meaning that interconnects a  
> stem resource with an ontolex:LexicalSense resource. Option two is to  
> create an >automatic connection that assigns all senses of the  
> ontolex:LexicalEntry instance - to which the ontolex:Form instances that  
> contain the stems belongs - >to the stem resources.  E.g. the instance  
> ex:stem_play is a part of ex:form_plays and should have the same senses  
> as ex:lex_play. As a result each stem >resource would have exactly the  
> same ontolex:LexicalSense assignments as its corresponding  
> ontolex:LexicalEntry resource.
There are counter-examples where this makes wrong predictions. Consider  
nominal incorporation as in Inuktitut:

qimiq n.abs. "dog"

qimi(q)-u- v. "to have a dog" (verb with incorporated noun)

qimiutuq
qimi(q)-u-tuq
dog-INC-3s.VPART
„(he) has a dog“

However, as the 3s.vpart morpheme also serves to mark nominalization, an  
alternative gloss is „dog owner“ (n.abs). If the noun is ergative, this is  
the only reading:

qimiutup
qimi(q)-u-tu(q)-(u)p
dog-INC-3s.VPART-ERG
"dog owner (erg.)"

The point is that there could be a lexical entry for qimiutuq as dog  
owner, and a morphological segmentation, but we should not infer that a  
dog owner is a dog nor the other way around. Meanings of stems and lexemes  
must be clearly distinguished at all times. In European languages, this  
pattern is less systematic, but we do have lexicalization processes where  
stem meaning and lexeme meaning are quite far apart. Think of Italian  
isolamento "insulation", via isolare "insulate" from isola "island". An  
island may be isolated, but not necessarily in insulation.
>
> Since the stem is the lexical-semantic core of all grammatical forms  
> that can be build from a lexical entry this method would save time since  
> the senses do >not need to be repeated for each stem resource. However,  
> there might be too many exceptions applying to this and result in wrong  
> implications. Option 1, >in contrast, might result in inconsistent  
> datasets with stem resources pointing to different senses than its  
> corresponding lexical entry. Comments from >people working with lexical  
> datasets sharing their experiences and practical needs for representing  
> the meanings of stems would be very helpful here.
I would in fact recommend to provide stem-level sense information only if  
it differs from lexical-entry level sense information, but it must be  
possible to do so. Responsibility for sense consistency is with the  
creator, not with the data model.
>
> 2) Describing the meanings of grammatical affixes contained in  
> ontolex:Form resources
>
> According to the existing OntoLex-Lemon specification an example was  
> created that applies to the usage recommendation of reusing the lexinfo  
> >vocabulary:
>
> ex:suffix_s a morph:AffixMorph ;
>
>  lexinfo:number lexinfo:plural .
>
> The statement that should be expressed is "The suffix -s encodes the  
> meaning of plural.". Within the discussion two possible interpretions of  
> the given >example emerged. A) It states that the meaning of the suffix  
> -s resource is plural. B)  It states that the suffix -s carries the  
> lexinfo number value of plural.
Not sure I understand the difference correctly, but lexinfo:number has a  
range restriction requiring that its objects are instances of  
lexinfo:Number. So, if anything, A) must be rephrased as follows:

A')  It states that the meaning of the suffix -s resource is the lexinfo  
number value of plural.
vs.
B)  It states that the suffix -s carries the lexinfo number value of  
plural.

With this reformulation, the difference seems to be that A') entails that  
there is no other meaning, whereas B) does permit that. In that  
interpretation, A') has counterexamples where a single morpheme expresses  
different functions at the same time: In Italian capi "heads", the  
morpheme -i expresses both plural number and masculine gender.

Best,
Christian
-- 
Prof. Dr. Christian Chiarcos
Applied Computational Linguistics
Johann Wolfgang Goethe Universität Frankfurt a. M.
60054 Frankfurt am Main, Germany

office: Robert-Mayer-Str. 11-15, #107
mail: chiarcos@informatik.uni-frankfurt.de
web: http://acoli.cs.uni-frankfurt.de
tel: +49-(0)69-798-22463
fax: +49-(0)69-798-28334