Re : Re: Universal Translation Language

Dans un courrier dat=E9 du 31/05/99 14:27:29  , Ray a =E9crit :

> Personally, I do not see how machines will manage translation of literatur=
e
>  unless they become truly intelligent. =20
the chief of the french automatic translation programme i interviewed=20
(Maurice Gross) was very sarcastic about the failure of the european=20
programme EUROTRA (which cost billions to eurotaxpayers) but quite confident=20
that within 30 years the main lexies of french and english will have been=20
listed and good translation made possible betwixt both languages. he defined=20
lexies as irreductible concepts expressed either by phonemes such as words,=20
compound words ("chair leg"), integrated PoS ("porte-manteau") or even=20
"fixed" sentence patterns ("ce n'est pas sans une certaine =E9motion que....=
"=20
or "do you want to come to my place for coffee ?) etc. he told me lexies mus=
t=20
be reckoned within sentences. so they are computing all sentences of some=20
daily newspapers for years and chase the lexies. he estimated the number of=20
basic lexies to some 100,000 and expected a similar number in both french an=
d=20
english. he told me : "i won't see myself the achievement of that task, but=20
you will" (he's 50 or so and i'm 30).

I know that in AI circles the
>  question of whether our present high-speed & very efficient morons will
>  ever achieve anything like real intelligence is the substance of argument=
s
>  & flames - and quite irrelevant on this list IMHO.
i feel that artlangers are best able to make an ial because they can write=20
poetry. poets know and mix words according to all shades of their complex=20
meaning. they instinctively know every meaning of each single word and all=20
meanings you can conjure up by pairing them.

Mathias