It really depends on what you are looking for. NLTK includes many
libraries and in most cases they are easy to customise for specific
needs/languages. I've used NLTK for a couple of years to analyse an
Italian corpus of spoken language. Obviously, it might be used as it
comes for particular tasks (e.g., statistic analysis with FreqDist
module), possibly easy to adapt to simple morphology (I'd suggest in
this case an integration with Morph-it, Baroni has worked on it and I
guess he is pretty confident with NLTK ) and impossible for semantic
analysis using WordNet (but note that WordNet clones for Italian are
available from the University of Pisa and the University of Trento).
You may start by looking at the rich documentation available at the
web-site:
http://www.nltk.org/documentation
If you have any specific question ask here or jusr feel free to send
me an email.