The best sentence segmentation tool until now?

I do not know about the actual best, but when I needed a sentence splitter fair few years ago, this was the best freely available one I could find then:

http://cogcomp.cs.illinois.edu/page/tools_view/2

It was not perfect but it was pretty good for my needs at the time.

More recently I have come across these:

• http://code.google.com/p/splitta/, which certainly makes impressive claims:

Includes proper tokenization and models for very high accuracy sentence boundary detection (English only for now). The models are trained from Wall Street Journal news combined with the Brown Corpus which is intended to be widely representative of written English. Error rates on test news data are near 0.25%.

• http://code.google.com/p/uima-sentence-splitter/

However, I have not had the chance to use either of them. I think GATE also has some sentence splitters, but I have no idea how good they are.

(Note: I didn't actually try any of these)

The splitter by Dan Gillick (splitta) mentioned in the answer by Daniel Mahler should be good, at least according to the paper describing it. However, as it is trained on WSJ+Brown, it might be overfitted and not the best for your domain.

Another alternative is the unsupervised algorithm by Kiss and Strunk (2006) which is implemented in nltk's "punkt" model. This implementation comes with a pre-trained English module, but can also be trained on any other (un-annotated) text.

The LingPipe Sentence Extractor works pretty well