[selllikesybok]

Unrelated to the OP, but what are you using to resolve place names to geography?

Seems to be mostly on target, but you get weird stuff like Camden Town (UK) -> Camden, New Jersey, US and it's common enough to notice.

[xando]

http://nlp.stanford.edu/software/classifier.shtml I'm not too happy about it.

But, I planning to replace it with simple knuth morris pratt text search against the list of all significant locations. This 100x slower but results are rarely bad.

[selllikesybok]

Cool, thanks for taking the time to answer.