Print

Print


> >this means the "semantics" of the machine translation
> >problem is buried inside the network of relations.
> >Does it really solve the problem?  I think a more
> >plausible way is to use so-called restricted input.
 
> This sounds interesting, but I have to admit I don't really understand what
> you're saying. Is it possible to put it in layman's?
 
the UNL does not provide a separate "?lexicon?" but instead
relied on english words to build a database, mapping english
words to foreign words.  that is the source of the potential
problem.  in other words, you get a tangled web of words
very difficult to organize or to develope further.  another
problem is that the UNL might even discourage unification
of languages by its emphasis on translation.  in my opinion,
an interlingua is the solution to language barriers;  but
the UNL doesnt really provide such a thing.  on the other
hand, i think the UNL project will benefit everyone if they
allow their results to be used by others.
 
> My question when I say the UNL page was to ask how this differed from any
> other interlingua approach ("interlingua" in the computer sense of course).
> Among the few things I know about machine translation (and am I even
> correct?) is that none of the approaches have gotten very far even after
> 40+ years, and that the whole endeavour has been a bit of a disappointment.
> So I wondered why the UNL page was so highly optimistic.
 
if you try to translate from one natural language to another
then it is extremely difficult.  but if the source document
is written in a restricted language,  there might be hope.
right now im reading to see what the current status is like?
- Yan King Yin