@kragen also, https://github.com/darius/sketchbook/tree/master/mutagen for a bit more text-gen stuff
@abecedarius Ooh, this could be interesting! Also, was the CMU Pronouncing Dictonary what you used for prosody in Trurl's Electronic Bard? The other night I did http://canonical.org/~kragen/sw/dev3/minpairsets.py after discovering that eSpeak is actually pretty reliable these days and made up a quick pronouncing dictionary with it, but a more reliable dictionary would make it work better. Last night I was also trying to figure out how to get internal rhyming to work (cf. Eminem's Without Me)
@kragen Good idea with the minpairsets! Yes, I used the CMU; I'll bet eSpeak or some such is a better start now.
But my rhyme/meter code may be helpful; don't think I've seen better elsewhere, though it's necessarily conservative. It ignores word boundaries, so internal rhymes shouldn't be hard?
@kragen I wonder what happens if you train a neural network poet on phonetics instead of text. Or I guess give it both.
@kragen P.S. I've been working on my español a little, although it's obviously too late for the eclipse.
@abecedarius I suspect eSpeak is less reliable than the CMU pronouncing dictionary. Ignoring word boundaries is probably a good idea, but reproducing family rhymes like this also needs some kind of fuzzier rhyme matching:
fˈiːl ðə tˈɛnʃən sˈuːn æz sˈʌmwʌn mˈɛnʃənz mˌiː
hˈɪɹz maɪ tˈɛn sˈɛnts
maɪ tˈuː sˈɛnts ɪz fɹˈiː
juː sˈɛnt fɔːɹ mˌiː
@kragen Yeah, when I said my rhyming was relatively good I didn't mean it shouldn't be easy to improve on.
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!