How does auto-correct work?

I use Swype on my tablet and have a bad habit of hitting psst before previewing. Auto correct has self corrected many words to something not even close to what I had intended. I swear if I ever meet the programmer who created my tablet’s auto correct I’m going to kick him right in the yamsack without saying a word.

I had to go back and change six words in this pis5 (there’s one, the word should have been “post”) that auto correct got wrong.

BAM! RIGHT IN THE YAMSACK!

The “learning” feature is particularly pernicious. Most people tend to make the same typos over & over. They also tend to use the same slang or jargon words over & over. Each time autocorrect helpfully changes the unknown word to something else and the undesired change isn’t caught by the user, it learns to be even more aggressive about making that change.

My phone makes a noise when it autocorrects. But half the time I’m texting in a noisy environment and never hear it. So even if I proofread carefully before sending and fix all the erroneous auto-corrects, I’ve already taught the darn thing to make the same bad corrections next time.

There ought to be a UI to clear out all the “learning”, or at least a way to identify and remove bogus “change *this *to that” entries in its logic.

My texts are loaded with work-related acronyms which autocorrect thinks are all typos. If I could tell it that “BYT” is a legit word but “byt” is probably byte or “but” that too would be a major improvement.

I know you can reset the user dictionary on an iPhone. I’d be surprised if there isn’t a way to do this on other phones.