Linearized computations - was OCR
Dr. Evil
drevil at sidereal.kz
Fri Aug 10 14:04:51 PDT 2001
> I think text is the wrong approach. Put up some pictures,
> and a question - 'Click on the white bunny' 'Click on the
> chipmunk' 'Click on the Russian icon'. 'Click on the clowns nose'
> 'Click on the clown's right hand'
>
> Taking text, which is explicitly designed to be readable, and
> trying to find a degree of distortion which is human (but not machine)
> readable is probably a bad approach. Try something a 6 year old
> can do, but which is still a PhD thesis problem for computers.
Actually, I thought about that option. There are two problems with
it: First, I need to find a bunch of non-copyright clip art which
displays clearly on any different browser, and then map it, and then
write a big database of questions. This isn't difficult, but it's
time-consuming. I'm a big believer in "do something now, refine it
later on".
The second problem with that is that for one of my applications, it
needs to be highly cross-cultural. Many of the users will be
non-English-speaking, who might not know what the word "clown" means,
and any non-abstract image has a potential for offending someone. If
one of my images is a pig, maybe some Orthodox Jews will be annoyed.
If one image is a woman with her face exposed, maybe some Moslems will
be offended. If one image is a policeman, maybe some c'punks will be
offended. Who knows. People are constantly seeking new ways of being
offended as a way of telling other people what to do. People should
have thicker skins, but they don't, and when it comes to making money,
I want customers, not controversy. The last thing I want is for some
rumor that my system is anti-something because some image of
something-or-other is used. As an example, for years there have been
rumors about Proctor and Gamble being run by Satanists, because they
have stars in their logo. Not goats, or a head with horns, but plain
old stars! A hundred-year-old well-established brand! In my effort
to support the largest user-base possible, numbers are the way to go.
(Oh wait, maybe I had better put a filter so it doesn't display 666 or
69 or 444 or 13...)
I'll make improvements as I go.
More information about the Testlist
mailing list