Share HN: I wrote an app to help my paralyzed cousin type significantly faster. | Hacker News Reader

Share HN: I wrote an app to help my paralyzed cousin type significantly faster.

chir.ag

240 pointschime29 comments

Submitter here. My cousin has not been doing too well for the past few months and so I put this project aside for now. However, a friend suggested that there might be many others who could benefit from such a tool or its derivatives.

This isn't a startup and I honestly don't want to treat it like a business. However, I do think it qualifies as a hack because of how it works. It uses Google Suggest in addition to a dictionary to try to guess the next word that the user is going to type. If you type WHERE, the next word is most probably going to be IS, THE, or DID. If you then select IS, the next word after WHERE IS will probably be THE or MY. KType uses this and a few such principles to make smarter, better guesses. It is not perfect but I intended to make it extremely smart and customized per user. When I was actively working on this, I had 100 ideas per day on how it could basically read the user's mind and suggest words and phrases most effectively.

My family and I looked at many existing systems for making typing easier but there wasn't anything out there so I decided to roll our own. As of right now, I don't know where to go with this because of my cousin's uncertain health condition. I would love to hear from medical and text hackers about developing this further.

Note: Currently I use Google Suggest localized for India so many of the suggestions may seem weird to non-Indian users. However, it's a simple change in URL to get localized suggestions for any country thanks to Google. Also, I wrote it for 1280x1024 resolution but it can be resized for most screen sizes.

This has great applications for mobile devices as well.

I think you should treat it like a business. Profit will maximize the impact of your change, and allow you to sustainably provide this solution to many more people than you would otherwise.

Google "Social Entrepreneurship" for more details.

Very cool. Feature requests:

- Accept keyboard input. Some users might find it easier to press a key rather than use a mouse or even a touch screen. It would enable support for stuff like this: http://www.usm.edu/ids/tlc/images/hw-ikeys.jpg

- Feature a big "copy text" button so that the user might be able to use your service and paste said text somewhere else on the web.

Those should be very easy to implement. Thanks!

This could actually go well with the CrunchPad.

I wish we would get more of these types of submissions on HN. Well done.

Indeed! Things like this are so inspirational

Are you aware of the dasher program? Because if you are not you really need to take a look at it.

It can let someone who is fully paralyzed, with just eye movement, type at a very good speed. And it works for less paralyzed people too (it has configurable modes).

I don't know why it's not more frequently used, but it should be - it's much better than most other typing software used in these cases.

Edit: I just noticed you listed it on the page.

Edit2: And I see everyone else told me that too.

This is the first thing listed in the "KType is not" section:

"KType is not a text-entry system for average users like T9 and Dasher."

Dasher was the first app we tried but it did not work for him because it requires a very high level of steadiness. My cousin's hand and eye is not steady enough. Plus you still have to type all the letters to spell any word. KType is not a replacement for Dasher but rather works at another layer (TCP vs HTTP) as it has more to do with guessing full words than typing individual letters.

You write that dasher is for the average person. I've only played with it, but the authors claim it's specifically for paralyzed people. I guess the claims don't live up to reality, it's quite disappointing.

There are different types of paralysis. Imagine a quadrapaledric who can control a pointer with his or her chin. This is a space where some tool is going to have a general success rate with each type of disability, but every case is going to ultimately come down to individual preference. Oh, and practice, practice, practice.

Dasher has a variety of options for how to drive it. Most of them are pretty forgiving of unsteadiness, and can type multiple letters per gesture. It isn't that great but some people with movement problems do use it, I think. (I typed this comment entirely in Dasher but I don't have any movement problems myself.)

Hi Chirag,

My cousin's child has severe Cerebral Palsy. He has a dedicated device[1] to help him interact.

Maybe you could include some ideas from these existing devices?

---

BTW, well done on your implementation.

---

I develop a mouse accelerator which makes moving the mouse a lot easier. When it is ready, I will post that onto HN. Your article has shown that the reception will be good.

It suits disabled people who have poor muscular control over the mouse.

------------

[1] MyTobii ::

http://www.inclusive.co.uk/cgibin/sh000001.cgi?REFPAGE=http%...

http://www.mytobiicommunity.com/

That's one very cool hack man, something to be proud of.

There are large bodies of text such as wikipedia and others that you can download, maybe you could use them to aid your prediction ?

It would also make the application work when offline.

I had originally planned on making this an offline app but scrapped that when I realized it would lose the ability to do real-time Google Suggest (or similar searches via Yahoo/Bing/Twitter). Search engines have a huge database that I can use to help with suggestions.

Even if I make an app for iPhone/iTouch, most users can still access the web and make use of Google Suggest so the effort to go through 20GB of Wikipedia/DBPedia/Freebase data and implement my own intelligent search didn't seem worth it.

Ok. If you ever want to go 'offline' let me know I'll be happy to contribute to your project, it looks like it would be a worthy way to spend some time.

Hey thats a good use of bigram probabilities :) Try reading in loads of text (Indian blogs if you want, but wikipedia would be more accessible) to build bigram probabilities, this way you can totally remove dependency on google suggest (and things people do not search for that often). I can help if you want :)

trigrams probably work even better, two words preceding gives a lot more certainty about the third.

yeah, but then you need orders of magnitude more data for building the dictionary which is orders of magnitude larger :)

I'd say going from two to three is `half' as more effort as going from one to two (in logarithmic scales). And that is precisely because trigrams can predict better --- they are even more redundant than bigrams.

This is very cool - stroke victims could benefit from this too.

Sorry for being dense, but although the clicking and prediction and word selection are just fine, I don't understand how this is used to actually type anything _into_ any place else? I couldn't select the typed text by dragging my mouse, or anything like that. How is this intended to be used?

Also: Interesting that you laid out the letters simply in alphabetical order (as opposed to QWERTY, Dvorak, or anything else).

> KType is currently only useful for typing in short phrases and hasn't been integrated into a large text-editing webapp yet.

It shouldn't be too difficult to incorporate this into a webapp that lets the user type lots of text - sort of like the T9 mode on your cellphone. The original goal was to allow the user to type about 140 characters (SMS/Twitter) and hit a Submit button.

The reason for choosing alphabetical layout was to minimize mistyping due to unsteady hands. Three rows like a QWERTY etc. keyboard means when user is trying to hit a key in the middle row, they could accidentally hit a key above or below the desired key. In two row layout, this issue is minimized. Also notice the keys are in a piano-keyboard style layout with ample gap around each key, thus reducing the chances of error further.

reminds me of the system that Prof. Stephen Hawking uses. Also it seems there would be some benefit of acceleration at the word and phrase level by using a system like the awesome bar. You want to analyse the persons own output and prioritise phrases based on that. You could also have a top phrase list - "where's my drink" "i need the toilet", etc. could be highly optimised that way.

This is the best hack that I've ever seen posted on Hacker News. It really serves as a reminder that making a difference in the world is about solving real problems for people. Solving real problems won't always make you rich, but they are just as important. Inspirational -- thanks my friend.

didn't work for me, but then i am using links2

This is awesome! Congrats... seriously well done.

Reply on news.ycombinator.com