View single post by Robert
 Posted: Mon Nov 26th, 2012 03:19
Robert



Joined: Mon Apr 2nd, 2012
Location: South Lakeland, UK
Posts: 4066
Status: 
Offline
Doug wrote:
Eric wrote:

The real answer is to read what you or the computer have written, before hitting POST. :-)


Exactly - if you type 'word' correctly it does not get auto corrected
if you type incorrectly and tap the 'x' it does not remember the misspelling
if a correction pops up in the middle of a word keep typing

Used properly it saves a lot of time

I think another big problem is hesitation and mistrust of the system (I do hands on with a lot of people and these users are the ones who have the most trouble)

Here are some example of iOS autocorrect working based on typing 'wor' correctly then mistyping the last letter

if you type wore then it does not autocorrect (perhaps later after you have taught iOS that wore is not a word - see note below)

worr, worf, worx, worz, wors, worg all become word

worc becomes worcester
worv, worb, worh become worn

worj, woro, worl become work

Note:
I deliberately typed worw which iOS tried to correct to word
I deliberately typed worw again which iOS again tried to correct to word
I deliberately typed worw again and iOS did not and never again will attempt to auto correct
I can fix this by resetting my spelling dictionary - you can work out how for yourself, this is iOS - not rocket surgery

If I just want iOS to forget the word worj then I must make iOS try to auto correct TO worj and tap the x then repeat
I did this by typing worq, tapping x, typing worq, tapping x again and now typing worw now autocorrects properly again

Note 2:
Sometimes typing worw autocorrects to different words (eg wore)
This is because iOS uses pattern recognition to calculate intent and increase the target size of some letters based on where your fingers tapped and in which direction for previously typed characters


Eric is right, of course!

However... Proof reading your own text is not the best way to tind errors, after all you KNOW what you typed and tend not to 'see' errors. The spell check finds spelling errors as a rule. Little red dits under the word, excellent. It can't find words which are correctly spelt bit out of context. If some logic were applied that has to be possible.

Being pretty deaf I do it all the time, during a conversation I assemble the words I do hear and understand, put them into context of the conversation and fill in the gaps. Where this does not work is with numbers and names I don't know. Otherwise I can hold a reasonable conversation by only hearing (and understanding) half of what the other person is saying.

All this presupposes that you look at the screen while you type? I am far too busy trying to find the keys for the letters I think I want to use, I glance at the screen every half dozen words or so to see what I have, going back and correcting the occasional word with the little red dots under them, IF I agree they are wrongly spelt.

So, I sometimes never know the auto replace 'feature' has done it's little deeds.

Learn touch typing I hear you scream! No way, I have no wish to become a touch typist, in fact I find the idea, well let's not go there...

My brain, if it's wired at all, is not wired for typing. Making things, yes, repairing things, yes but repetitive automation definitely NOT.

So my stance is that the auto completion idea is on the right lines but has some way to go. I have read that Siri uses some form of logic which may improve the auto completion feature, if so, it needs implementing in the auto completion feature. In the mean time I am torn between turning it off to stop it putting silly words in my text and being a Luddite and turning it off.

I think for iPad users it's a different situation because they are already looking at the screen, even if they like me have to hunt for the right keys.

This is a good example of what I don't like about autocomplete, HOW does it come up with THIS!!! Why did it not simply flag it as wrongly spelt luddite???

Attachment: Screen Shot 2012-11-26 at 08.09.27.jpg (Downloaded 9 times)



____________________
Robert.