Dans un courrier dat=E9 du 28/05/99 21:18:39 , Marcos a =E9crit :
> >To sum up my post : tagging adjectives, adverbs, participles, etc. is (i)=20
not=20
>=20
> >necessary for human speech=20
> =20
> Of course is not. But it's highly recommendable in an unambiguous
> language.
>=20
i don't get that one. did you ever read the Chinese "white horse" paradigma =
?
=20
> >and (ii) not enough at all for translation=20
> >programmes.
> =20
> Of course not, again. Some syntax rules are needed as well.
> =20
yeah. some thousand ones :-)
> >You may rather want to precise what word or clause refers to what former=20
or=20
>=20
> >latter word or clause, then what semes of each word or clause are=20
combined.=20
>=20
> >But deciding one word only works in a sentence as an "adjective" or an=20
> >"adverb" or even a "noun" or a "verb" is just illusory : a word is=20
> >"adjective" to a noun but may be "adverb" to the verb integrated into=20
that=20
> >noun ("good fisher") and reversely, etc.
> =20
> That's what PoS marking and syntax rules are for in an unambiguous
> language, to determine what word/phrase/clause affects which
> word/phrase/clause, and how.
>
if you succeed in pairing semes within and between each of such PoS and stil=
l=20
are able to speak "that" then i'll be the first one to learn your language.
The only thing i know for sure about all that is what the guy in charge of=20
the french automatic translation programme told me once : words are no=20
problem, the problem is multi-word, semi-integrated lexies which are the=20
nucleus of natural languages and are a hundred times as many as words.
you think making compound or derived words out of two "independent" words or=20
morphemes is benign, but it's not because this multiplies semantic=20
integration links with different words in the sentence.
> =20
> =20
> Saludos,
> Marcos
Salut gars !
Mathias