[KLUG Members] human vs bot
Adam Tauno Williams
adam at morrison-ind.com
Tue Apr 12 14:20:08 EDT 2005
On Tue, 2005-04-12 at 12:08 +0200, Bert Bbbink wrote:
> > Does anybody know why these things are so hard for an OCR program to
> > handle? I know the human eye has a much easier time with recognizing
> > characters than a computer does, but with all the hardware we can throw
> > at a problem, do a few lines scratched through the figure really confuse
> > a computer that much? I'm just imagining the screaming that will occur
> > when somebody figures out a simple algorithm to strip out the lines.
> you can raed this lien of text vrey well I tink. Humans read sentences by
> recognizing the words mainly by the fisrt and lsat character.
> Bots need to "read" every character seperatly. Most ocr (as far as I know)
> find letter by trying to lay them on numerious images of characters.
Right, thats exactly the difference. The OCR programs aren't 'reading'
at all. A person understands that a line is one thing and a character
is something else, they can overlap on the page but you almost can't
even glance at that page without realizing the distinction. Even the
best OCR program is a very very long ways from that.
Everyone I've seen can barely interpret characters on a blank white
page. The first part of the sales pitch is chest-thumping confidence,
the second part (after the demo) - a long tedious explanation of why
this that and the other thing came out resembling a random string of
characters.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://www.kalamazoolinux.org/pipermail/members/attachments/20050412/3c0ce5d1/attachment.bin
More information about the Members
mailing list