W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > October to December 2003

Re: Captcha and inaccessiblity

From: Tom Croucher <tcroucher@netalleynetworks.com>
Date: Sat, 18 Oct 2003 15:26:16 +0100
To: "w3c-wai-gl@w3.org" <w3c-wai-gl@w3.org>
Message-ID: <oprw8s12jxu930jj@mail.icet.co.uk>

I forget to mention the address of the CMU project http://www.captcha.net/

On Sat, 18 Oct 2003 15:18:23 +0100, Tom Croucher 
<tcroucher@netalleynetworks.com> wrote:

> Captcha stands for "Telling Computers and Humans Apart (Automatically)". 
> It most commonly involves using an image with warped text in it to 
> 'verify' that it is a human and not a robot submitting a form on the 
> internet. However the problem is that currently all the systems rely on 
> putting a piece of information in form which computers cannot interprete 
> and asking a human to repeat it back. This is becoming more widely used, 
> yahoo uses it to verify mail being sent, network solutions use it to stop 
> third parties piggybacking on their whois lookup system. With spammers 
> and other service abusers so prevelent the big service providers are keen 
> to protect themselves but obviously this has impacts on accessibility.
> I spoke to the guys at CMU about some stuff I was interested in with text 
> based captchas a while back (6mo maybe) and they said they were working 
> on a text version. This would probably take the form of a question which 
> is common knowledge or analogic. ie, "If I have five apples and I eat 
> three. I give one to Bob but Susan gives me two apples, how many apples 
> do I have?". To which the acceptable answers would be "3", "three", "3 
> apples", "three apples". Another example might be like those IQ tests "A 
> dingy is to ship as a go-cart is to a?". The acceptable answers being 
> "Car", "Bus", "Lorry" etc.
> The other obvious way to go would be audio, so generate a word in an 
> audio format or some other sound. So, "What animal makes this sound?" 
> stuff like that. Which is almost inevitabley easier than finding a way to 
> generate textual questions which are understandable enough to humans, but 
> obscure enough to Natural Language Processors and Cognitive engines, 
> without being so long and convulted they are obnoxious to use.
> Some vague thoughts and possible solutions. I would love to hear ideas 
> people have on ways we could make this work without the currently 
> inaccessible images. I think we should insist that if people must use 
> image Captcha they either also provide audio ones or a phone number that 
> people can use to a bypass code.
> Tom
Received on Saturday, 18 October 2003 10:26:21 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 16 January 2018 15:33:46 UTC