hello, I'm an undergrad student doing a project on measures on semantic similarity.
I was wondering if anyone had any experience in using / comparing the WordNet Similarity measures. My main issue is trying to scale them to a common scale- preferably a scale of 0-1 with 1 being the highest correlation of similarity. however, with lesk and jcn, I'm not quite sure how to do this - lesk seeming to be on a scale of 0 to infinity. I noticed that a paper by Jarmasz and Szpakowicz used results from the Jiang and Conrath algorithm, scaled to 0-1, but they don't mention how they did this. If anyone has run into this before at all or has any ideas, they would be most welcome. (and apologies in advance, hope this is coherent, has been a late one!) Danielle.
