Optimising the T9 Keyboard | Hacker News Reader

Optimising the T9 Keyboard

torvaney.github.io

45 pointstorvaney17 comments

Awesome research! I was just thinking the other day, about how a qwerty layout for T9 might have been easier and more efficient. Cool to see something even better. A little off-topic, but this 6-button keyboard format looks perfect for my dream device.

I take a lot of walks. I often listen to an audio book while I do so. I also frequently get texts, or hear something in the book I may want to write down.

I've always wanted to be able to handle these situations without taking out my phone, but rather clicking buttons in my pockets. For example, I get a text while listening to music. Click a button to hear it read out. Type out my response. Send it. Open notes, write a reminder to buy bread. One-handed, blindly, while on a walk outside. No voice recognition - no "OK Google" nonsense.

I've considered getting a Twiddler - but they're expensive, and I don't think I can easily configure it to do these things - particularly browsing and responding to messages - or opening up a new note, while listening to an audiobook.

Does anyone know some combination of software and hardware that can do some of this for me?

I’d suggest to revisit voice recognition — it works quite well for me in the same usecase.

I also like to take walks, sometimes listening to podcasts. The stock iOS voice recognition (the microphone button on the keyboard, not Siri) is quite good, I usually just talk into the phone without looking at the output. After the walk, I format and clean up the notes to fix any errors.

That's fair - I live in a bustling city and get self conscious with people being able to hear me text and such. I know it's really effective these days though.

I used to be able to write stuff blind pretty reliably back when I was using a Palm device. Its “Graffiti” system was pretty good. https://en.wikipedia.org/wiki/Graffiti_(Palm_OS) A keyboard based on it could go a long way.

what you describe sounds a bit like https://www.tapwithus.com/ -- but i have never used it and i suspect it's more of a novelty than a genuine utility.

I remember back when I could ride a bike while texting and still watch the road. T9 was the high point of text input on handheld devices!

This one is fun. Maybe one optimisation worth checking is what's possible if you have to stay alphabetical.

E.g. T9 puts 4 letters on '9', presumably a good choice as 'X' and 'Y' are rare. Would it be worth it to shift the 'V' from '8' to '9'. '3' has 'D' and 'E' on the same key, maybe move the 'D' to '2' ?

Yes, it's interesting that the T9 seems to have made a slight improvement on my "naive" alphabetical layout. You might even be able to brute-force a solution where the alphabetical constraint is imposed.

From what I can remember from using a t9 keyboard quite a long time ago, was that you pressed each key to cycle through the letters on the key. with a break if you need multiple letters on the same key after each other.

the whole predictive word thing was something pretty much alien to me.

Pretty cool read. My biggest problem with T9word is Spelling. If you misspell something on a full keyboard, the other person may be able to ges by sounding it out. T9word would throw out a word that would be no where near what you want. 'Ges' would become 'her' and the other user would never think 'guess' was the intended word.

My 2nd biggest problem with T9word is you need to look at the screen to verify the word you wanted is the one guessed. The other t9 keyboards allowed the user to type without looking at the screen. (Such as under the desk during class )

Absolutely. This will be an even bigger issue for people texting in a second or third language (which I understand is quite common among countries and communities where feature phone use is high).

As I understand it, one of the big advantages of T9 had over other, more sophisticated forms of predictive text is that each word in the dictionary can be encoded in close to 1 byte. Given the constraints faced at the time, T9 feels to me like it is close to a local optimimum.

My idea was instead to do the combinations that you will used in punch cards in computer, and you can push together multiple keys like they have multiple holes in one column, but no actual cards are needed for this method.

(Note: I have never used the computer with punch cards)

I dislike touch screen and voice controls, so using physical keys is better (whether T9 or other methods).

This is great! I like your site and would like to continue getting updates from it. Would you consider adding an RSS feed?

I tend to blog more often on my soccer blog (https://www.statsandsnakeoil.com), which has RSS set up. Naturally, that's a bit more niche. I'll look into RSS for this site if I start adding more new posts and projects.

I don't think this is very insightful. Using the first half of books for training data and the second half for testing data is still training the model specifically for these texts and authors. Not quite as bad as testing on the training data, but not great.

Hello! OP, here. I agree that the task is training for this specific subset of English writing, which isn't ideal.

For this task, I was primarily interested in whether the task would work at all. My assumption is that given we can optimise for these texts, we could optimise for more representative datasets, too. Perhaps you think this is a weak assumption?

Do you think testing on a sample of totally different texts from different authors would be more convincing?

Reply on news.ycombinator.com