Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

You're right, but it's not just hardware.

I was part of the spell correction team at Google, and we made sure that we aren't overly aggressive even though lots of people mistype their web query.

Nowdays I find it much harder to research rare things on Google, and I have to undo the automatic correction that the spell corrector does all the time (which is OK as long as it's easy to undo).



I've always preferred the results page corrections for spelling.

I'm more ok with the extra step, if it's my fault for misspelling.

Less so if I have to take an extra step to correct Google's inaccurate "correction."


A question for you, given your expertise: why do most (all?) spell checking/correcting not take into account key locality (is, neighboring key accidentally pressed). A big one for me is hitting n or b instead of space. I think it has gotten better recently, but has a loooong way to go. Just curious on your thoughts. Cheers!


It was there from the first iteration of spelling a long time ago (the first iteration was just looking at misspelling frequencies, second iteration 12 years ago moved to understanding multiple spelling mistakes/mistypes)

If you see these kind of easy-to-correct mistakes, it probably means that the spell corrector wasn't given that big attention in the past 10 years.


I just realized that you wrote space....that’s harder as the spell corrector works on words separately :(

I was working on URL corrections part that didn’t need to use segmentation, so I don’t know the details though.


Swiftkey does.




Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: