Consider harvesting short sentences from parallel data

[gregtatum]

It's not guaranteed that parallel "sentences" are actually sentences. We could write a script to detect how many sentences are in each parallel datum, and then attempt to extract them into a new dataset. We could use word alignments to ensure that the left and right sides of the sentence agree on being part of the same sentence.

Perhaps this won't work, because there aren't many multiple sentences in the datasets, but it seems worth looking into.