SydLexia.com - The Forums :: View topic

Posted: Sep 09 2010 10:56 am

Captcha is, of course, the annoying verification system most sites use during registration. The format is usually two randoms generated words that you must type in. If the people running the site are morons, they'll make the text case sensitive and ridiculously hard to read. Captcha is a pain in the ass, but it's necessary to keep spambots off of sites. And every so often you'll get a funny one. Here's one of my favorites:

Coffin Exchange sounds like something Tom Bosley should be advertising during commercial breaks on The Price Is Right.

"When your loved one's life insurance doesn't cover the casket they deserve, Coffin Exchange is there. We work with Medicare to get the coffin you want at no cost to you!"

That, or it could be an emo band.

Posted: Sep 09 2010 10:59 am

Hah, nice topic. I can't remember any offhand, but know I have seem some pretty awkward ones. Great avatar btw.

Title: Titlating Joined: Dec 17 2007 Posts: 5042

why are you doing this and not making comics

Posted: Sep 09 2010 01:52 pm

Somehow I succeeded in typing this out on the first try.

Posted: Sep 09 2010 01:57 pm

What the fuck is that second word?

Posted: Sep 09 2010 02:10 pm

Something in Hebrew. I had to guess what it was in Character Map. And I had to do it backwards because of the way it's written.

Posted: Sep 09 2010 03:10 pm

I thought that looked like Hebrew, and Coffin Exchange sounds more like a Magic card.

Posted: Sep 09 2010 03:18 pm

lordsathien wrote:

I thought that looked like Hebrew, and Coffin Exchange sounds more like a Magic card.

Hah, it does sound like an emo band though.

Joined: Dec 28 2009 Location: New Jersey Posts: 66

Andrew Man wrote:

lordsathien wrote:

I thought that looked like Hebrew, and Coffin Exchange sounds more like a Magic card.

Hah, it does sound like an emo band though.

Or a dating website for undertakers.

Joined: Dec 06 2007 Location: Connecticut Posts: 11244

The-Excel wrote:

Andrew Man wrote:

The-Excel wrote:

Somehow I succeeded in typing this out on the first try.

What the fuck is that second word?

Something in Hebrew. I had to guess what it was in Character Map. And I had to do it backwards because of the way it's written.

I am almost certain it is the name:
Alexander

From right to left it reads אלכסנדר

Of course there are no vowels, so it is really guessing from _lchsndr. The _ is silent but usually indicates that a vowel would be placed on it. Also the "ch" is that sound you always hear in Hebrew that doesn't exist in English. Like Ch in the word Chanukkah, Challah or Chutzpah.

For what it is worth, Google agrees with me.
http://translate.google.com/#en|iw|alexander

2011 SNES Champ Joined: Dec 04 2006 Posts: 1929

ReCaptcha always shows you a word you need to type out correctly and some other word it doesn't give a shit about, so you can write whatever you want and it'll still accept it.

Posted: Sep 09 2010 07:26 pm

So whose idea was it to contribute Hebrew text into a Captcha pool?

Joined: Dec 06 2007 Location: Connecticut Posts: 11244

The-Excel wrote:

So whose idea was it to contribute Hebrew text into a Captcha pool?

The process is automated.

The text is derived from automatically scanned shit and then pushed one word at a time. Over a long period of time machines can learn from the results and be better at predicting text out of images. This improves OCR technology.

My guess is that this Hebrew word was contained inside a document primarily written in English.

http://www.google.com/recaptcha/digitizing

Posted: Sep 09 2010 11:27 pm

With application that sophisticated, you'd think there would be some kind of reporting mechanism for things that do not belong in Captchas for English-language websites.

Joined: Dec 06 2007 Location: Connecticut Posts: 11244

The-Excel wrote:

With application that sophisticated, you'd think there would be some kind of reporting mechanism for things that do not belong in Captchas for English-language websites.

The point is that it takes an image of scanned words on a page, it then breaks them down by the spaces in between the words. Then it pushes each word out to see what type of answers it gets.

No one pre-OCR scans the words. They are simply located on the sheet and passed on.

In a recaptcha window one of the words has been regularly guessed and has a high success rate where people are able to type the same response for the image. The other word has a lower success rate, or perhaps is a new entry. In this case, the word will continue to have a low success rate until it is flagged and a person looks to see why. They will eventually see it is not English and will pull it.