2 Apr
2022
2 Apr
'22
9:55 p.m.
[it's unfortunate to lose track of the value around pursuing the task, so it to pick in what way to pursue it.] [what if we were to just produce the output, and not try to find any return? we could add just one or two extra things to try to increase the return, maybe. how would you produce the output in the simplest way possible?] [i guess i'd use the human transcriptions as a train/test set, and then add heuristics to improve the output if needed, in feedback, as considered before] [well, that's really similar to the original plan, so maybe we should go ahead with the heuristics.] [ok i think. i think i was getting confused around processing the way the current model transcribes letters and symbols as similar-sounding words.]