removes exceptions from splitters

[asdqwex]

Not sure why that is in there but it looks like it is unneeded. Feel free to merge or reject as you see fit.

[coveralls]

Coverage remained the same at 100.0% when pulling bf45b5a on asdqwex:master into 4880754 on jsvine:master.

[coveralls]

Coverage remained the same at 100.0% when pulling bf45b5a on asdqwex:master into 4880754 on jsvine:master.

[jsvine]

Thanks @asdqwex! If I recall correctly — and it's been a while since I wrote those lines — the exceptions are meant to keep certain period-separated abbreviations stuck together, e.g., so "C.I.A." remains "C.I.A." and not three sentences ("C", "I", "A"). Ideally, markovify would have a more complete list of these abbreviations ... or a more sophisticated/generalized approach to detecting non-sentence-ending-periods. What do you think?

[alpha-beta-soup]

Shouldn't it just consider any letter/s-separated-by-periods as acronyms, and not break them? U.S.A. and USA are semantically the same thing. I can't think of many cases where any period that is not followed by a whitespace character is sentence-ending. You just have to be careful to note that not all acronyms are \b(?:[a-zA-Z]\.){2,}, but also \b(?:(?:[a-zA-Z])+\.){2,} (the latter matches Ph.D whereas the former does not).

[jsvine]

That sounds like a promising approach. Thank you for suggesting it.

[jsvine]

Very belatedly, thank you for the suggestion. v0.8.3 now checks for a generic pattern, rather than a preformed list.