Some undersegmentation of sentences

[nschneid]

Despite manual editing there are still places where a long sentence ought to be split up (esp. where it consists of a blockquoted sentence with > followed by a response). Looking for multiple consecutive spaces in the raw text uncovers some of these (as well as some discourse appendages like emoticons, which should probably remain in the same UD sentence).

It would be nice to write a script to help clean these up—the tricky part is updating offsets in each parse.