<div dir="ltr"><br><br><div class="gmail_quote">On Wed, Oct 1, 2008 at 8:57 PM, Joel Newkirk <span dir="ltr"><<a href="mailto:freerunner@newkirk.us">freerunner@newkirk.us</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>
<br>
On Wed, 1 Oct 2008 20:32:34 -0600, "Ori Pessach" <<a href="mailto:opessach@gmail.com">opessach@gmail.com</a>><br>
wrote:<br>
<div class="Ih2E3d"><br>
> I have a question about the "predictive" keyboard shown in one of those<br>
> movies - what on earth is the justification for providing the following<br>
> predictions - User input is on the left, prediction is on the left:<br>
><br>
> w - a<br>
> wo - so<br>
> wor - wot<br>
> worl - work<br>
> world - ? (I'm not sure what was the prediction, but at this point it's<br>
> moot... It was probably wrong, anyway.)<br>
><br>
> The qtopia keyboards do the same braindead thing. They're perfectly<br>
> unusable<br>
> as a result. The handwriting recognition engine appears to be wired into<br>
> the<br>
> same "prediction" engine, resulting in letters that were recognized<br>
> correctly to be randomly replaced with incorrect letters.<br>
><br>
> Is this a bug, or is that the intended behavior?<br>
><br>
> --Ori Pessach<br>
<br>
</div>Intended, you just have to understand what it's doing.<br>
<br>
Unlike T9 and other predictive inputs that are familiar from cellphones,<br>
the keyboard predictive process looks at neighboring keys, 'predicting'<br>
that periodically you may hit the wrong one. (if you're using your finger<br>
that's pretty likely, especially if they're as wide as mine)<br>
<br>
If you pay close attention, the actual user input is "helko worlc", and yes<br>
it suggests 'hello' and 'world' as the prime suspects.<br>
<br>
As a developer and hacker I hate it. For text messaging it seems it would<br>
be useful, presuming that the user won't always have (or choose to use) a<br>
stylus.<br>
<div></div></blockquote><div><br>I understand what it's doing. It's not doing it well. I tried it for shell input, and it was an unusable mess. I tried it for text messaging, and it was an unusable mess. It has no model of the likelihood of erroneous input (relatively low) and instead appears to look for the word with the closest minimum edit distance to the user's input. This is nuts. I have never - literally - gotten the word I typed in. In the common use case, of a user who enters a correct word, it invariably get it wrong.<br>
<br>Understanding what it's doing doesn't make it less of a nuisance.<br><br>--Ori Pessach <br></div></div><br></div>