What can be done is to use cluster analysis (k-means)

on the semantic network. Instead of the Euklidean

distance, use the semantic distance. Each cluster

will represent one meaning.

How to get the semantic distance: create a graph

from the Wordnet database. Using Dijkstra to

obtain the distance between two words is not

the best choice. Better do a random walk and

take the logarithm of number of visits as

distance measure.

Regards,

Joachim

Hi Marouane,

I've used wordnet on R. The WordNet library contains the function synset which give you the synset for a word, i.e.,  all words with the same meaning. For polysemic words, you'll get a synset for each meaning.

Here you have a nice web for understand the wordnet structure.

http://elies.rediris.es/elies9/2-4-2.htm

Regards,

Mari Luz