Predictive text is great for composing text messages. But it has problems. In particular: paragrams (or textonyms, or whatever). Because each key represents 3, or even 4, letters, various combinations of keypresses can result in numerous different possible words.
The combination 63 could produce "me" and "of," 5477 could be "kiss" or "lips," and, an extreme example, 22737 can produce 14 different words. Try it.
I propose a solution. Or, at least, something that could significantly reduce the likeliness of accidentally typing "shag" when you mean "rich" — with my name, this happens a lot.
Built-in grammar checking. Not of Microsoft Office proportions, obviously — that's both unnecessary, and probably technically unfeasible. Any phone with predictive text already has a fairly extensive dictionary of words. I don't suppose it would be too difficult to attach to each word what part of speech it is.
Example: in the case of me and of. "Me" is an objective pronoun, and "of" is a preposition. So, if I were to write a text saying "Don't you love me," it would easily tell "me" was what I intended, rather than "of" because "love" is a verb, and an objective pronoun functions as the target of a verb.
Obviously, it's not without technicalities. "Love" can also be a noun, and the preposition "of" could perfectly naturally follow it. So it would need to see "you love" and see that "love" is being used as a verb.
Actually, just while writing this, I've realised how horribly complicated it could get. But I reckon if you had every word in the dictionary tagged with what part of speech it can represent, and give the software the ability to not only compute every possible word, but every possible grammatical arrangement, it could eliminate a large percentage of paragram issues.
There might not be much point if you can't eliminate all of them, you could argue. But Microsoft Word doesn't even come close to being able to eliminate all grammatical errors (in fact, Word's is one of the worst built-in grammar checkers). It would be a fair trade for the problems it could fix.
Just a brainstorm, anyway. What do you think?
» Sorry Camino, I Went Back To My Ex-Browser's Edgy Twin ... Last Reply: 1 year ago by alexsuraci.
Mike, a millisecond (ms) is a thousandth of a second, not a hundredth.
The first WebKit nightly time I saw in that list was 3861.2ms, which is nearer 4 seconds than 37 seconds. I suggest you re-run your numbers. Because 88 seconds is abysmal — I'm assuming you got somewhere in the region of 8,800ms, right? That makes for 8.8 seconds, not 88.
I got 15,552ms (15.5 seconds) in Camino, for what it's worth. On a MacBook with only 512mb RAM, without bothering to close "extraneous" apps or restart the browser after having it open for over 24 hours. (And I've experienced none of the problems you describe, Mike.)