Hello all,
 
I haven't joined in on the subsequent discussion because classes have started again here and I haven't had time to digest all of your comments.
 
I still don't have time - except to clarify a point. I didn't mean to say that the actual name of the relation was significant (although perhaps that is what I wrote) - I meant that the semantics of it was. That is, the combination of the intended meaning of the name and the possible values of the relation (is it a boolean property, an otherwise two-valued property, or a multi-valued property), that has to be carefully chosen.
 
As for the problem with 'before' (or 'after') needing two different kinds of values, boolean as well as the two objects exhibiting precedence, I grant you that it's not what you want. My way around it would be to only have one relation, let's say 'before' and place the arguments accordingly : A before B vs B before A. 
 
But obviously there are more things to be said in response to the entire discussion. I just can't get around to that now.
 
Greetings,
 
Lyne
 
------
Lyne Da Sylva
Professeure agrégée
Responsable du certificat en Gestion d'information numérique (GIN)
École de bibliothéconomie et des sciences de l'information
Université de Montréal
C.P. 6128, succursale Centre-ville
Montréal, QC, H3C 3J7
(514) 343-6444 (tél.)
(514) 343-5753 (fax)
[log in to unmask] 
http://mapageweb.umontreal.ca/dasylval
 

________________________________

De: Word Grammar de la part de And Rosta
Date: dim. 07-09-08 07:14
À: [log in to unmask]
Objet : Re: [WG] RE : [WG] Default inheritance and relations



Some responses to Dick & Lyne:

> Da Sylva Lyne wrote:
>> Hello all,
>> 
>> Here's a (humble) shot at it :
>> ("it" being : Precisely what is it that prevents some word W from inheriting the following?
>> [12: W subject Z]
>> [13: W after Z]
>> [14: W before Z]
>> I've had various thoughts, but none that I really like, so I'd be interested to hear other ideas.)
>> 
>> The only reason that you know that W cannot inherit both [13: W
>> after Z] and [14: W before Z] is that you know that they are
>> incompatible, i.e. that 'after' and 'before' are two opposite
>> values of the same relation of precedence. You could not infer that
>> from, let's say, names like 'rel1' and 'rel2' if you did not know
>> what these labels meant.

I agree with Lyne completely: the key thing is that part of our knowledge of 'before' and 'after' is that X can't be both before and after Y.

('Before' and 'after' aren't the best example, bcs IMO they're synonyms (converses), tho part of our knowledge of this relation is that if X is before Y then Y is not before X.)

>> Here's another example : if your two relations were 'head-of' and
>> 'dependent-of', one object obviously could not inherit both [A
>> dependent-of B] and [A head-of B]. But it's because you know the
>> semantics of the relations that you can infer this.
>>  
> ## This is where I start feeling uncomfortable about this solution.
> Precisely which relations are mutually exclusive? This example is
> difficult because I believe there are cases where two words may be
> mutually dependent; e.g. in /Who came?/, /came/ is the complement of
> /who/ and /who/ is the subject of /came/ - and 'complement' and
> 'subject' both is-a 'dependent'.

The relations that are mutually exclusive are those that are incompatible through logic or stipulation.

>> Incidentally : I actually have argued extensively in my thesis that the choice of the names of relations used in grammatical representations, as well as their values, have to be chosen carefully based on an analysis of the nature of the linguistic phenomena involved - that relations often exhibit oppositions of various nature which suggest what the appropriate formal representation should be. This was based on markedness theory. (It also fits in quite nicely with the notion of default, as I argued that unmarked values are unmarked precisely because they are default values of the oppositions in question.) For example, I claim that 'before' and 'after' are not the proper representations for phenomena of precedence, since this exhibits a binary opposition of a single relation (precedence) which is best represented as either a boolean-valued relation or some other two-valued relation.
>>  
> ## In contrast, I think I would follow Sydney Lamb in arguing that names
> should do no work at all in the analysis, because all the information in
> a network is carried by the way the nodes are connected. But this is
> probably a separate issue.

I too wd insist on the insignificance of names. I'd go slightly further than Dick in that I see the difference between converses like 'head'/'dependent' and 'before'/'after' as an insignificant matter of naming.

> ## My idea is this. The crucial difference between my two examples
> (about locomotion in birds and word order) is that in the case of word
> order, the two potentially inheritable relations (before and after) both
> apply to the same argument and the same value. Because of this, the
> inheritor has to be more fussy about conflicting relations so it avoids
> inheriting sister relations (a kind of incest avoidance, if you like),
> whereas in the other case it just avoids inheriting two examples of the
> same relation.

But what other examples of 'incest avoidance' are there? There's no problem with X being both colleague of Y and friend of Y. There is no principle or effect of incest avoidance.

--And.