k-punk

Not sure if its the same thing but a lot of phones in the US have what they call T9. Doesn't quite predict the way you describe, but its very similar. It only generates words the same character length as you type though. What it does is have you press each number only once and then guesses the letter you want to used based on probability and a stored dictionary. For instance typing my name (ABE = 223) generates instead the word "BAD", the most common word generated by the 223 sequence. Quite clever really although it generates some funny shit sometimes: http://www.collisiondetection.net/mt/archives/000892.html

Yes the T9 is what I use and what most phones have here. It is very useful and easy. The best thing it does (not of much relevance right now) is transpose "Xmas" as "Wobs", a much better and entirely secular name for it. The worst thing it does is CONTINUALLY type "rub" instead of "pub" despite "pub" being plainly the all-time most common word I use in txts.

sadly and curiously it is not programmed to included TXTSPK!! this = daft imo

the groups of equivalence are fascinating and repay study

i can of course see why xmas mayn't be included in a chip's kosher dictionary but WHEREFORE THEREFORE "WOBS" as the korrekt alternative? it isn't the beginning of a real word surely? is it just the little program giving up in lynnetrussian despair?

Oh also the T9 program is a friend to anti-rockist drinking viz. "shot" over "pint".

That's true, but between 'shot' and 'pint' comes 'riot'.

The reason this is good is of course because it makes texting at least 50% faster, aiding telecommuncicational banter, incl. international banter. Great!

It told me everything about the state of Britain that my phone could predict "SATC" but it couldn't predict "colliery". Also fine with "Big Brother", but not "Orwellian".

Disappointing.

Worse yet, try typing my name in with it on!

(I do use it, though.)

yeah and ms word does it too.... like when you type "human" it tries to write "human resources"

I've lost count of how many people have said to me that predictive texting is shit & how they never use it then once you've shown them how it works and how much quicker and better than all that txt msg shorthand bullshit, they never look back. The rhyme or reason to its suggestions is simply a dictionary database!

txt-spk obv > predictive txts bcz it saves #.

yeah predictive texts actually improves peoples spelling grammar etc cos it is much harder to right 'nite', 'gr8', even 'c u l8ter' or what veer than write their full blown equivalents. a mobile in every classroom etc!

A flippant post has actually yielded some very interesting material ----

seems to me that even when you can fathom out its arcane methodology - and my friend Aled today gave me some instructions that make it less opaque - predictive texting is retteritorializing and negative because as Julian so pithily shows, it takes up more space than txtspk and, in suppressing txtspk, it inhibits linguistic innovation (some wd call this 'improving ppl's spelling and grammar').

The whole politics of anticipative text replacements v. interesting (esp. when they are done by Bill Gates on Word)...

I think it uses Bayesian analysis and a big dictionary, or more likely Bayesian analysis of huge wodges of public domain text which then creates a dictionary.

I hated predictive for ages and some versions of it still don't work well enough. But I understand how it works now :-). Wouldn't be without it.

Anyway, the proposition that evolving mobile technology ellides txtspk seems a bit conservative to me. The point of txtspk is that it has invaded most forms of written media and thereby created new forms of discourse. The fact that mobiles are now an unhome for it (unless you choose to select a menu item) might not be that much of an issue.

I got a new Nokia 6230 yesterday -- it's FANTASTIC! It's got a radio in it!

Txtspeak is the lingua franca of IM, for instance.

What's bayesian analysis Paul?

What's IM, Tom?

I think the story goes like this. Thomas (IIRC) Bayes was a 16th century monk who figured out how, in theory, you could statistically analyse large volumes of data to identify patterns and make predictions. There were some freaky algorithms involved, and it all remained theoretical until computers got fast enough. The formal definition is:

"A statistical approach that takes prior information into account in the determination of probabilites. The Bayesian approach yields an equation that must be used to convert P values obtained from Chi-squared analysis of recombination data into actual probabilities of linkage between two loci."

In other words, we know X just happened; how does that affect the probability of the next event being Y? For example, we know based on statistical analysis of English texts that if you type PQRS on your phone, you're probably going to be using a word starting with S; type DEF, and you're probably thinking of a word beginning RE; type MNO, and you probably meant PEN, or one of the ten or so alternatives in the "Matches" listing. It dynamically changes the probability assessment as it goes.

Bayesian analyis gets used for search engines, spam filtering, medicine, all kinds of stuff. It was really hot when Autonomy floated for a billion quid (= hypothetical whole company valuation of a billion) a few years back.

Wow, that's really interesting Paul (sounds sarcastic but I'm genuine...)

do you work in comptuters then?

I fink predictive txt iz gr8 i culdnt txt wifout it lyk!! it savez so much time!!

nope...sorry.dont use it...yes...we all got marked for grammer at school...rite spelling and all...but u know...people like 2 abberviate or shorten or use their own words...i think i'm that small crowd that LIKES to spell stuff out 4 themselves..t9 can kiss my ass....cause i dont use it...refuse 2 use it...and abuse ppl when they use my phone and use it and DONT change it back...hahahaha....