Is there any web-service that provides a word semantic-similarity measurements?
I'm aware of Disco, but I'd prefer a service with an ongoing growing base (and it would be most helpful if you had tried it in your projects).
I'm also aware of WordNet-based algorithms, but installing and managing it as part of the project resources brings heavy weight.
You could consider using the google index API, there is a sample application here - http://www.mechanicalcinderella.com/ uses it and has links to the original paper describing the technique. Source code here - http://complearn.org/index.html.
i have not used it myself, other than for fun... but it seems pretty decent.