i should make a page that's like "RISC-V extension or randomly generated bottom keymashing"
@whitequark Hmm, let me slap an underbar on those and run them through c++filt
@whitequark thinking of training a "small language model" on these
@whitequark i have no idea what i'm doing and i like how one of them is just "duct"
@dramforever @whitequark out of curiosity, what specifically are you training there? My nickname was generated with a homemade genetic algo, pronouncability score, negative score for matching existing words, and a dictionary file. And looking at way too many results until something clicked. Wondering if there was a smarter approach I missed;)
@psentee @whitequark i trained an hmmlearn.hmm.CategoricalHMM on the assumption that it would be able to learn patterns like "prefix", "middle", "end of string", used my lucky seed 3 for training, and tweaked the number of states to see how it would behave
@psentee @whitequark (there's not enough data for it to give anything that makes any sense, but that's the fun part, it's not like the original makes any sense to begin with)