This problem has been addressed in psychology/psycholinguistics in the form of feature norm experiments, where participants are asked to list defining features for given words.
Data from some of these experiments is available online. See for example http://ppw.kuleuven.be/concat/workshop2007.php
Some people who have run such experiments: Ken McRae and Gabriella Vigliocco for English, and Sabine Schulte im Walde and Alissa Melinger for German.
All the best, Katrin
On Feb 1, 2008 10:35 AM, saket <saketn at gmail.com> wrote:
> I'm looking for some advice on how to approach the following problem.
> I'm trying to create a "feature vector" for a list of common objects.
> For example, for a "banana" I'd like to be able to say that it has an
> "elongated SIZE", "yellow COLOR", "smooth TEXTURE", etc. This could
> amount to looking up the object in the dictionary or WordNet and
> extracting adjectives and noun phrases, but I also need to be able to
> automatically infer that adjective1 corresponds to the descriptive
> feature for "size", adjective2 for "color", etc. This could also be a
> probabilistic construction, which says that it's "usually yellow, but
> maybe also green".
> Essentially, I'm trying to create a database defining objects based on
> their descriptive features, which I think can be used for a variety of
> tasks. Have people approached this problem before? Does there exist a
> corpora which might be relevant here? What's the best dictionary to
> use? Any advice?
> Saket Navlakha
> saket at cs.umd.edu
> Corpora mailing list
> Corpora at uib.no