A programmer called Cyrus Lopez has published a collection of every accidental 17 syllable haiku’s from the Reddit comment directory dataset, dating from December 2005 to October 2016. The result is 349,162 haiku’s at a total of 5,519,927 words.
Essential to haiku is Kiru, or ‘cutting’ usually achieved with a ‘kireji’ or a ‘cutting word’. This is the the slicing of a poem’s content into halves; the offset of a primary description with a disruptive secondary. To take a popular example from Kobayahsi Issa:
Earns his living
As the composition of the individual poems from the Reddit dataset is effectively random, kireji is rarely employed. However, in the act of extracting 17 syllable from 1.5 terabytes of JSON code, effectively slicing haiku into purely metric analysis- perhaps the project retains a certain shock.
The highest scoring haiku from public vote on the website is currently Haiku #231708:
Can’t you see how much
better you make the world just
by being in it?
You can vote for your favorite here.