Re: Unambiguous languages (was: EU allumettes)

* Mark P. Line said on 2004-05-07 23:26:24 +0200
> Responding to taliesin and Philip here on this topic. (I'll respond to
> And Rosta's suggested ambiguous sentence after I've reviewed what I
> used to know about CY scope semantics.)
>
> taliesin the storyteller said:
> > * And Rosta said on 2004-05-06 05:02:33 +0200
> >>
> >> I may have mixed CY up with another conlang in my memory, but my
> >> recollection is that it adopts Word Net as the inventory of its
> >> word senses. By so doing, it gets rid of ambiguity arising from
> >> polysemy.
> >
> > It gets rid of *English* polysemy.
>
> What other kind of polysemy does it *fail* to get rid of, or what
> other kind of polysemy does it (unintentionally) introduce?
Uhm, you do know that polysemy is wholely dependent on the language it
occurs in, right? If the world's languages divided the world up in the
same categories, Machine Translation would have been a lot further along
by now. I'm not saying there isn't an overlap, I'm not saying it adds
ambiguity, all I'm saying is that it makes the divisions a slightly
abstracted English makes, and that other languages makes other
divisions.

> What difference does it make what (real or engineered) language is
> taken as the starting point as long as (a) the resulting lexicon is
> unambiguous, and (b) the semantic scope of the resulting lexicon is
> broad and deep enough to support at least the same range of
> communication as a natlang would
You'd need a deeper/broader Net than Princeton WordNet for that. Have
you had a look at CyC?

Don't think that I am against WordNet in any way, it's great! I've
actually started to map the meanings of words in Taruven to WordNet
meanings, and it is interesting to see what WN covers and what it
doesn't...

> (other than the fact that it's not sexy to start from English)?
You might already know this but curious onlookers might not:
EuroWordNet is a collection of Word Nets for european languages. They
have not attempted to map non-english languages onto the Princeton
WordNet as, for instance the French refuse to divide their language up
into (American) English categories. I'm not making this up. The
EuroWordNet was funded by the EU...

I haven't studied this sub-EuroWN in depth but it's a fun toy:
<http://sisx03.si.ehu.es/tresnak/wwni/index.htm>

A less ridiculous reason for not using the original WN is that it makes
up several more or less arbitrary (not words original to me, now where
is *that* paper...) words to "fill in" the tree and make it almost
binary and the resulting graph does not necessarily adequately encode
tha language "as she is spoke". The Dutch WordNet for instance, is a lot
flatter... let me find that paper...

<http://glahn.hf.ntnu.no/~torbjorn/undervisning/ling3304/elra.rtf> Have
a look at the picture on page 4.


t.

Reply