segment s splits s into a list of sentences.

It looks for punctuation characters that indicate an end-of-sentence and tries to ignore some uses of puncuation which do not correspond to ends of sentences

It's a good idea to view the source code to this module, especially the test suite.

I imagine this sort of task is actually ambiguous and that you actually won't be able to write an exact segmenter.

It may be a good idea to go see the literature on how to do segmentation right, maybe implement something which returns the N most probable segmentations instead.