Re: Nouns, verbs, adjectives... and why they're pointless
From: | Joshua Shinavier <jshinavi@...> |
Date: | Thursday, December 10, 1998, 12:15 |
> (Such a challenge ... here is how I think of it, sortof:)
> Nouns are objects of thought. Verbs are relations between them.
> Adjectives are attributes, like one-sided relations.
> Everything else is built from those. Danoven seems to use
> levels of speech instead of POS, but either way one builds
> the same parse-tree. Thinking in terms of "argument valency",
> nouns = 0, verbs/preps/conjs 2, adjectives/adverbs 1.
There are also intransitive verbs, with valency 1, but this is basically
true -- which is one of the main problems with the system; nouns are not
chosen because they have a valency of 0, but because they seem "static"
(many of them, anyway...) and physical (again, many but by no means all of
them), and the 0 valency is *imposed* upon them. And a maximum valency of
2 is imposed on verbs, although further objects are simulated using
prepositions (e.g. "HE gave IT to HER"); objects of nouns are simulated in
the same way (e.g. for an object: "a FRIEND of ANIMALS", for a subject:
"the SINGING of ANGELS" -- annoyingly the subject and object of a noun are
grammatically indistinguishable).
A better idea is to leave PoS and valency unrelated, allowing each word its
own appropriate valency rather than restricting it to that of its word class.
In D. even a "noun" or "adjective" may have a valency of 4 (t3 -- a subject
and three objects). PoS distinctions restrict the form of the conceptual
trees which may be built; to me, to have an idea in my head and to not be able
to say it because my language doesn't like it is extremely aggravating...
aggravating enough to make me want to go to the trouble of creating a better
one, even ;-)
JJS