Things I'd Change About Auto-Correct
techreflect.net
techreflect.net
Seriously, this isn't even a feature for anyone ever, it's a straight-up bug I seem to get hit with several times a day.
But my absolute biggest problem with autocorrect isn't even on this list.
It's when I swipe-type a word and the correct word comes out. Then I swipe-type a second word, and my iPhone decides the first word was wrong, and changes it to be more probable with the second word.
Now, it's super-frustrating but wouldn't be the end of the world if I could just easily tap somewhere to restore the original first word. Occasionally it's listed as one of the options at the top of the keyboard, but not usually. Usually I've got to backspace through both entire words, retype the first word again, hit space twice so the iPhone "forgets" about being in double-word-correction mode, hit delete twice to delete the period-space, hit space again, and finally type my second word!
It drives me nuts. I read what I'm typing as I type it. I don't ever want my phone to fix a word earlier. I wish so badly there were an option to disable this.
Personally I use /s/old/new which also had the benefit of some platforms, like slack, automatically fixing the original message.
s/search/replace
Bothers me that it won't trim a trailing / tho.FWIW it's usually faster just to press up (another shortcut to edit the last message) and word jump to fix.
Also FWIW I've always seen the asterisk apended.
“appended*”
Why have we accepted this reality? That's what's truly bizarre.
Slack can do it, what are they so special?
I guess maybe in a fluent back and forward it might slow people a little worrying about things they shouldn't. But this is not my experience.
This inability seems the IT industry at it's worse. It's like some Frankenstein putting immutability onto human beings.
I don't understand. Communication has historically not been possible to "edit". There's nothing inhuman about that. If I say something, I've said it. Of course I can later make a correction, but the words I said remain said. The same has been true for the thousands of years we've had written communication, until a few years ago (on some platforms).
Personally, I find editing messages a misfeature. There's few things more annoying than replying to a message, and then finding out that the message you're replying to has been replaced by another one. If a platform has this feature, it makes me less likely to use it.
Have you never had an argument get to 'you said this', 'no I did not' ? We even edit our own messaging in our heads.
We make mistakes, why can't we correct them?
(We are talking autocomplete of spelling and grammar here at the original level)
And since obviously your end can have a setting/hack/default to keep the original anyway I'm not sure what the fear is.
Spoken, yes. Written, no. But sure, most communication is (and certainly was) spoken, I'll give you that.
> Have you never had an argument get to 'you said this', 'no I did not' ? We even edit our own messaging in our heads.
I'm afraid I have. They suck tremendously. Do you think these arguments are a good thing?
> We make mistakes, why can't we correct them?
We can. We post a correction. "Sorry, I meant to say this."
> (We are talking autocomplete of spelling and grammar here at the original level)
Unless you produce a magic AI that lets people fix only spelling mistakes and nothing else, we're not talking about just that.
> And since obviously your end can have a setting/hack/default to keep the original anyway I'm not sure what the fear is.
Most communication platforms these days are actually not controlled by the users, so no, I can't have a setting for it. And if I have a hack for it (which probably requires a tremendous amount of technical expertise, out of reach for most people), I'm almost certainly in violation of some EOL and may get banned. Heck, in some crazy jurisdictions I could even end up in trouble with the law.
You said “9:00”.
“go check”.
and that for a simple interaction, imagine a complex argument...
Plus it's silly, most mark the message as changed and some have an audit trail like Slack.
I'm not sure what scenario this plays out in?
As far as an audit trail is concerned, you'd need to have all the cell carriers update their SMS infrastructure to save the associated edit history along with the text of each message. That's definitely non-trivial.
I'm not saying it's a bad idea to have this feature, just that it's much more work than you realize, and the benefit is probably not there to justify the work. Slack has the luxury of controlling their entire platform. They don't rely on Ericsson or Qualcomm or Nokia to implement new features the way SMS would.
I mean, do I really have to reread all the "marked as changed" messages even though only a tiny part may have changed etc...?
Normal people tend to fare very well with "sorry, made a mistake: ...."
Handles a lot of these cases:
* Accidental last row instead of space - handled
* Easy undo - first backspace hit returns to explicitly typed out text then next one backspaces
* Whole word flashes on autocorrect
* Current candidate for autocorrect is centered on screen above keyboard
* Valid word won't autocorrect (can't actually test this since both devices are probably fully trained on what I type)
In my experience, my Android phone replaces valid words all the time to my great annoyance.
A couple of minutes ago I tweeted a post that included the word somehow. GBoard changed it to *someone. I went back and highlighted the word, changed it to somehow, verified it was corrected and submitted. In some fraction of Google's wisdom, it recorrected the word back to someone. Argh
Autocorrect is a mess.
In Italian "c'è" means "there is", which is supercommon and also a pain to get right with apostrophe and diacritic, but "ce" is a valid pronoun: if you start typing by typing "ce" it will autocorrect to c'è, but if you go on typing "l'hai" ("ce l'hai" = "do you have it") it will autocorrect the "c'è" back to "ce".
For some common words it doesn't do it, but for most it does. One common and consistent example is mot (towards, against) which becomes möt (imperative form of meet)
The problem is the cross-over between spell checking and finger location.
if I mistype something because my finger comes down on an adjacent key, with autocorrect turned off, I can figure it out.
Say I want to type QWER and type FDSA by mistake, tomorrow when I read what I typed, figure out where my fingers were and recreate QWER.
But with autocorrect which does spell checking and semantic stuff, there's a good chance that FDSA will be corrected to something like FORD or VISA and I will never be able to recreate what I was thinking of.
I'm sorry I don't have a less contrived example.
The reason I did this was because I had some important things I jotted down quickly, and later - even with an hour of thinking - I could not recreate my original thought.
also...
I do end up with n instead of space all the time.
stuff like thenquicknbrownnfox happens often.
And plenty of blame should be reserved for the keyboard design itself, causing mistakes. I literally started developing a new keyboard extension as a side project for no reason other than to move keys that constantly tripped me up. (Worst offenders are the stupidly-close backspace and action buttons so that half the time I send wrong text when I meant to delete a letter.)
I have a high end iPhone and autocorrect routinely slips in some corrections after I've already scanned the text visually and decided to press send.
The corrections are usually not desired, and often replace valid sentences/words with guesses that are more common words, even at times words or abbreviations that I know I have texted years ago but never since.
I'm not sure what's worse, the bad behavior of AutoCorrect or how laggy it is on my iPhone XS.
It's a tough one, though the answer really relies upon setting a probabilistic threshold.
The answer to many of these problems should lie with 'user settings' somewhere.
One thing I miss about Windows Phone keyboard was how good the prediction was. On iPhone it can predicts a useful word from time to time after months, but on WP I could type parts of long sentence without having to enter any new letter.
I've noticed my friends who use auto-correct usually don't compound words, which can change the meaning of the sentence completely.
In Finnish, words can have over a thousand inflected forms. Granted most of them are not used usually but every day I bump into a word that the autocorrect knows but when I add the inflection, it no longer recognises it. It tells me that it doesn't actually parse or "know" Finnish text, it just has a huge word list.
The second issue is that it decides which language's suggestion to show based on the characters I've typed (since I have it set to both English and Finnish). Normally it shows suggestions from both languages but if I type a single ä, ö, or å, it never suggests an English word. This is unfortunate since I sometimes hit such a character by accident and I have to go back and retype the word from that point to get back the English suggestions.
The only problem i have now is with the i not becoming capitalized automatically.
I'd also take a truly fixed and useable autocorrect solution over a larger Frankenstein camera any gooey applause riddled release event of a decade.
Should be easy to recognize.
Depends on the keyboard of course, but there really isn't a good override by default.