peb

🐭 πŸ•β€πŸ¦Ί πŸ”ž

disabled chimera thing
unabashedly weird & kinky
doing its best
owner πŸ’›


webbed site
pebble.pet/
fedi / mastodon
critter.cafe/@pebble
πŸ“
Baltimore MD

hey nerds help me out

i'm trying to figure out the logic for a function that'd detect a keymash within a message and it's proving harder than i thought

i know i don't want to use the dictionary approach of figuring out what isn't a keymash bc that'd be too much to run for every message

i was thinking maybe like detecting ~5 vowels in a row but that'd not work for some keymashes and prolly work for some obscure words, i'm kinda lost tbh


You must log in to comment.

in reply to @peb's post:

if we're to avoid using ML entirely for this, I'd suggest trying to get rules based on the likely position of the person's hands on the keyboard and tendencies to slide fingers across the keys, but this seems like it'd still be really complicated. it could be used more simply to simulate keymashes maybe, and then you'd have something to train a simple classifier that could tell you how confident it is that it's looking at a keymash, with a dictionary being compressed into the parts that know it's not a keymash, I guess...

oh no you're making me want to train an ERT for this, I have too many projects

that is like super complicated and way outside the scope of my abilities haha, but if you wanna take a shot at it for fun feel free to :D i need something simple even if it'll get a false positive every now and then both bc complex stuff would take me way too long for what this is for and bc my raspberry would prolly die

yeah sorry I just look at this problem and go "oh no this is one of those problems", it's like teaching a computer to read letters on a page, it's harder than it looks

if I give my approach a go it won't be for months, I have shaders to do and drawings to shake out of my brain

Pinned Tags