Re: distributional knowledge representation
Rob Freeman wrote:
> In an earlier post I mentioned work on the indefinability of "meaning" by
> Adam
> Kilgarriff at ITRI Brighton. I said I wasn't sure if he was still working
> with distributional approaches to knowledge representation.
>
> It seems he is. If anyone is interested they can check out some of his
> more
> recent work here:
>
> www.sketchengine.co.uk
The sketch engine seems to have some value in identifying and tabulating
some properties of words in a corpus. Thanks for the URL, Rob.
> Also along these lines is the following distributed "concept discovery"
> system, which may be of interest.
>
> www.leximancer.com
>
> -Rob Freeman
The concept engine is a beautiful presentation, but how is that helpful when
building an NLP application? I can see how it helps a human operator to
cluster similar concepts in a graphical way, but is there something that
I'm missing about how to mechanize that "similarity" in a NLP Q&A system?
Nice pointers,
Rich