[Developers] wordnet?

[Joel Truher]

The YAGO approach is indeed what I had in mind, thanks for the link, I
wasn't aware of it.

To me there seem to be two steps, though: (1) mirroring the external
resource, and (2) semantically integrating it.

I agree with crism that (2) isn't worth attacking comprehensively, but
I think there's value in (1), and (2) can proceed slowly, piecewise,
and still be useful.

But I'm a beginner.  Are there issues I should learn about with
respect to separating the process in that way?


Joel


On Mon, Dec 1, 2008 at 1:21 PM, Ed Summers <ehs at pobox.com> wrote:
> On Mon, Dec 1, 2008 at 2:44 PM, Christopher R. Maden <crism at metaweb.com> wrote:
>> The problem that keeps stumping[*] people who take this on is that
>> Freebase is language-independent, while WordNet is inherently
>> linguistic.  The English word "dog" can't just be attached to the
>> Freebase topic for dog-the-animal; the first synset for "dog" in WordNet
>> can be *connected* to it, but shouldn't be overlapped with it.  There
>> are things that can be said about "dog" that are not true of "chien" in
>> French.
>
> Yes, another way of looking at it is that as you traverse link between
> pools of data on the web-of-data you will encounter semantic-drift
> [1]. The drift between some data set will be more than others.
>
> I guess I just wanted to point out that the links were there if Joel
> wanted to try using them. The YAGO folks have an interesting paper
> that discusses how enhancing wikipedia data with linkages to wordnet
> can prove useful.
>
> //Ed
>
> [1] http://en.wikipedia.org/wiki/Semantic_change
> [2] http://www.mpi-inf.mpg.de/~suchanek/publications/www2007.pdf
> _______________________________________________
> Developers mailing list
> Developers at freebase.com
> http://lists.freebase.com/mailman/listinfo/developers
>