dissection in the concrete garage
the water with the labyrinth
unholy in our latent symbols
he found a snow were slowly gathering
on the smooth concrete garage
the disordered machinery at dissection
emerged from that of the night
so secretive and the aperture these monstrous structures
far from the sunken star-born
in acid the terrible receding city
multicellular protoplasmic bubbles faintly audible
and infinitely distant wastes
the madness admitting that voice
at last for almost be in many
at the previously perhaps that in
had seen what ails me as
the human madness as it as our aerial survey did
every fragment whose ranks shot suddenly as that accursed realm
damn em often that the highest peaks black morass
aug 3, supervised generation from bigram models, texts: Lovecraft: At the Mountains of Madness, The Call of Cthulhu, The Case of Charles Dexter Ward, generator: epogees.
OK! we were talking earlier about how much work the poet does vs how much work the reader does in constructing meaning.
so just to try it out, I generated the first 2 stanzas with a particular narrative in mind: I generated 3 sets of 14 lines, and picked phrases that would support the narrative; I also selected “next words” consistent with the bigram model, a couple times. for the last stanza I pretty much picked what came out: I generated 4 lines of 6 accented vowels and 4 lines of 9 accented vowels, then removed 1 from each group. oh yeah, also the beam search was selecting for “AE” phonemes; you know, like “AAAAAAAAH!!!” lol but the point is I didn’t particularly “author” the last stanza. so hopefully it will be more “open” in terms of allowing multiple types of readings.