this stinks of bad localization.
I suggest a DHTML "drop a cube on the square hole" test.
The text part of Captcha will always be Epic FAIL.
This looks like a promising CAPTCHA approach. I disagree with Chad because the text part allows it to be accessible to visually impaired users. The use of free semantics should pose a sufficiently difficult challenge to machines.
I think the text part is essential to make CAPTCHA accessible and fair to blind users. Also, computers arguably have more difficulty understanding world semantics than recognising images, so I think image CAPTCHAs will eventually be overtaken by semantic ones like this.
The accessibility for blind is a very valid point.
I know that computer vision can recognize images, so I wouldn't use them in such a simple state. I would look to 3d constructs and alternating perspectives to confuse the computer.
Basically, a physical Turing test, for the "submit" action.
A physical Turing test sounds interesting, but I can't imagine how it can be made accessible though.
Whether webmasters would adopt an accessible CAPTCHA is another issue. If they do, then I'd expect CAPTCHAs like Egglue to be widespread in the near future.
This honestly looks easier to hack than other captcha tests. Unlike turning images or sounds into words, which are human percpetions, grammar has logical rules.
It looks like it wouldn't be too hard to write a bot for this, even without a dictionary list.
I hate that the answers to the "bot" question all involve more work for the user, i.e. any type of captcha. Can't we focus more on just eliminating bots? I don't know enough about them to give suggestions. Off the top of my head, couldn't we register our computers IP with our ISP and have that validation automatically opt us out of having to fill in CAPTCHAs ever? If you choose not to register because you are scared of big brother knowing what you are doing and tracking it OR if you are a bot then you are subject to these types of CAPTCHA validation tests. Just throwing the thought out there. I hate to see the innocent user have to append their flow to accommodate jerk off bots.
I'm trying it out now and it indeed looks quite promising. Responses are only correct if I give a verb matching both the subject and the object. So it seems to check for probable truth of the statements, which I think computers generally have a hard time in understanding it.



Comments