portmanteaux random diastic
outh nows broked bosome withou youthou
unsweet outh yournity thould wortue eterfumentle
evisary lents theeking resensull thiness wortue
imped mistain wring Musick
4/26/13, Method 8e39ddee-aece-11e2-b790-f23c91aec05e (portmanteaux random diastic). Source text: Shakespeare sonnets.
Character-level markov generation is great because you come up with some awesome portmanteaux. Basically, halfway through forming a word, the system “forgets” the context that was used to generate the first part of the word, and starts generating a new word. This is particularly effective with n-gram models where n is around 4; more than that and you have too much context, less than that and you get some long weird words.
In this variation, I use charNG to identify portmanteaux, then use eDiastic with solid-vowel “words” to read through, to create a rhythm to the verse.
Method 8e39ddee-aece-11e2-b790-f23c91aec05e (portmanteaux random diastic).
Step 1: use charNG to generate from character 4-gram models.
Step 2: Use charNG to analyze portmanteaux.
Step 3: insert portmanteaux into eDiastic for authoring. For a seed text, use a vowel pattern that sounds good. “Read through” in random order, cycling.