Okay, I am certainly not the first to notice some issues with the auto-correct software in iOS devices. But what I find especially puzzling are some of the decisions it makes.
This afternoon I'm typing the words "to ensure" on my iPad and, as is often the case, I somehow miss the space bar, leaving the two words together - e.g. toensure. The iPad then, as now, offers to replace it and, because I type relatively quickly on the iPad, it happens before I can make a decision about it.
The word it selects? Tonsure.
Now, I fancy myself a relatively well read and educated person, with a fairly decent vocabulary. Still, here I find that I am - not for the first time, mind you - having to look up the word that iOS has elected to replace my text.
Incidentally, tonsure refers to the shaved crown of a monk or priest's head. Look it up yourself if you like but, honestly, why would I lie about this?
When you do look it up, you'll find it to be a word first encountered in the 14th Century, derived from old French and Latin. As I gaze at this in the definition I wonder once again what algorithmic logic went into making the device assume that it was somehow more likely that I wanted to make reference to a friar's bald pate than simply having accidentally tied two correctly spelled words together.
I write on the iPad often, and while I dearly love my device, this feature is routinely the most frustrating aspect of using it. Among the things I write are psychological reports, and the word "pattern" features prominently in them. For a long period of time I had problems with hitting the "o" instead of the "p", which the iPad would then helpfully correct into "oat tern", apparently believing that I was writing about grain farming near the sea. Every time I add an additional "s" to the end of the word "was" I get "Wasserstein". What's more, while it offers these miscorrections routinely, it continues to resolutely leave the word "fir" in place each time I type it accidentally instead of "for", thinking either that I'm writing a newsletter for the American Evergreen Society or perhaps that I am writing dialogue in a western novel.
What I don't get here is how it makes those decisions. It clearly has some capacity to learn - I don't see "oat tern" any more, despite my continued errors with the word "pattern" - but it seems to make poor decisions in regard to the likelihood of a given word. What is the likelihood that I actually was trying to write "tonsure", a middle-ages relic of a word instead of, well, anything else? While I appreciate that these words are in the dictionary on the device - and that I learned a new and interesting word today - it would be great if Apple would perhaps re-think the words it offers based upon some calculation of the statistical probability of the word being correct.
Another alternative would be to allow the autocorrect to be turned off, while leaving the spell check turned on (this does not currently appear to be an option). Because the autocorrect puts in real words for relatively fast virtual typists the end results are sometimes rather strange, and more challenging to detect because, though the words are wrong, they are spelled correctly. This option would be helpful tonsure the errors would be caught when the piece Wasserstein proofread.