NLP in general is very useful so you might want to broaden your search to general application of text analytics. I used NLTK to aid MOSS 2010 by generating file taxonomy by extracting concept maps. It worked really well. It doesn't take long before files start to cluster in useful ways.
Often times to understand text analytics you have to think in tangents to the ways you are used to thinking. For example, text analytics is extremely useful to discovery. Most people, though, don't even know what the difference is between search and discovery. If you read up on those subjects you will likely "discover" ways in which you might want to put NLTK to work.
Also, consider your world view of text files without NLTK. You have a bunch of random length strings separated by whitespace and punctuation. Some of the punctuation changes how it is used such as the period (which is also a decimal point and a postfix marker for an abbreviation.) With NLTK you get words and more to the point you get parts of speech. Now you have a handle on the content. Use NLTK to discover the concepts and actions in the document. Use NLTK to get at the "meaning" of the document. Meaning in this case refers to the essencial relationships in the document.
It is a good thing to be curious about NLTK. Text Analytics is set to breakout in a big way in the next few years. Those who understand it will be better suited to take advantage of the new opportunities better.
