Re: THEORY: Deriving adjectives from nouns

Dans un courrier dat=E9 du 05/06/99 09:02:04  , Tom a =E9crit :

> "Nicely dancer" makes no sense to me.  The -ly suffix there,
>  for me, could only be an adverbial usage, not the derivational
>  adjective suffix, as in "homely" and such (because "nice" is already
>  an adjective, and why do you need to adjectivalize what is already
>  an adjective?).
>
that's exactly what charles says : (i) this makes no sense in english and=20
(ii) this would be adverbial if it existed. the point is a very basic one :

a substantive may be an implied actor through integration (danc-er) or=20
semantic connotation (hammer). "dancer" has two possible "handles" : one as=20
"person", another as "to dance"; "hammer" as "item" and as "to hammer". some=20
words have only one "handle" : "end" is always an actor, "me" is always an=20
item (precisely because it always refers to an actor). even "mouth" has=20
implied semantic roles in many situations : speak, eat, yawn, gag, kiss,=20
drink, etc. it's a question of semantic "sp=E9cialisation" as we say here (b=
ut=20
apparently not in the US). now, when designing a loglang, you can't prune al=
l=20
those little ennoying "handles" poking out of the nice little sterilized bag=20
loglangers put arguments in to isolate them from predicates. so there are 2=20
solutions :
1. you create 40 or so relators to "absorb" the role power of the arguments;
2. you try to use as many words as possible like "hammer" or "dancer" with a=20
very specialized semantic role, whether derived like "dancer" or not like=20
"hammer". this is exactly what john cowan did a post ago with "pretty" : he=20
picked another word with a specialised semantic field.=20

that's what lawyers do all the time in order to precise whether they speak o=
f=20
the substance or the purpose (use or result) of a word ("fit for purpose").=20
this is not making things disambiguous, it's rather picking one of the two=20
following valencies of a single word : either the substance or the role. thi=
s=20
would be useless to do that with words having many possible semantic roles=20
like "tree", "mouth", "wooden stick", "flour", etc., because pointing to the=
m=20
as an actor just points so many possible meanings. but pointing to either=20
"dance-" or "-er" within "dancer" or between "fish-" and '-ing" in "fishing"=20
is very useful.
christophe does this with one of his conlangs so each word is suffixed a tag=20
showing whether it is used as an actor or an item. verbs are "acting"=20
substantives. it's very poetic and ambiguous. what does "mouth-actor" mean ?=20
this depends on the context. i like it very much. tunu does that too, but=20
only with paired words. this is used a lot in tunuan poetry.

>  > Precision is good, but so is ease of use.
>  > It seems every word is an encoded sentence.
>=20
i agree with charles. each word carries with itself whole experiences of=20
which it is an actor, from core cases to improbable white-man roles that mak=
e=20
you laugh when you get the joke. some may never get the point however.
=20
>  I don't think it's so much a matter of ease of use, as that a certain
>  level of ambiguity is actually desirable in a language.  Aside from the
>  fact that *no* ambiguity imposes what I believe to be a false dichotomy
>  on reality, it's practicly important to be able to convet concepts which=20
are
>  inherently ambiguous.  So, having a lexicon with no general word for
>  "animal", but thousands of individual words for individual species, seems
>  like an implicit denial of any interrelationship between those species,=20
> which
>  is clearly wrong, because all animals are obviously more closely related
>  both to other animals and to every animal in some fashion or another.
>  You could go even further, saying that there is a specific name for every
>  entity, or a name for every atom... you see where this is leading.
> =20
semes, sememes and classemes are used to select information within a concept=
.=20

>  The same thing goes for the grammar.  To encode every possible shade
>  of verbal mode, for example, into verbs by some system would seem to
>  deny any interrelationship between those modes.  "Desiring" and "wishing"
>  certainly seem similar to me.  But moreover, why can't you define a=20
specific
>  mode that means "wanting to do something while you are in your car becomi=
ng
>  irritated by traffic" (or something absurd like that)?  Not only do those
>  examples clearly seem, to me at any rate, to have certain underlying
>  relationships, just like animals are underlyingly related to one another,=20
> but
>  there is, a priori, no end to the kind of complexity you would have to ha=
ve
>  to have complete disambiguation of all forms.
>=20
indeed. but i believe human beings live a life where items and roles are=20
mapped according to certain habits. "banana" is likely to be grown and eaten=20
and "fabric" to cover, be worn or hang somewhere. same with roles : 50 roles=20
or so are most frequently used. complexity is not messy otherwise it wouldn'=
t=20
work.

i'm confident that very trivial experiences like "eating" or "wending my way=
"=20
or "i've seen that somewhere already" are fundamental roles rather than loft=
y=20
concepts like "agent" or "subject".
=20
>  So, the Golden Mean seems here to be the best rule to follow when
>  considering how much you want to build into your language.  (What
>  exactly that is, of course, is your personal choice, which is what makes
>  it all the more fun, no?)
some of us find it funny to call roles "dextro-illative case" and the like,=20
others to call them "to eat", "to feel", etc., and another ones to have none=20
of them at all.

mathias