On Captchas, and a nifty slide for those who can read Braille visually

[pne]

There's a slide show called Accessibility Problems with Visual Verification Systems talking about visual verification systems such as captchas (those images of distorted text that are supposed to prove you're human) and problems with them.

Many slides have images of typical captchas (e.g. those provided by LiveJournal, PayPal, Passport, ...); one particular slide had a "type the text of this image into this box" which probably not many people could comply with :) (to illustrate the point, I suppose).

I'll tag fledchen, leora, and pthalogreen as probably being able to.

Comments

[darth-spacey.livejournal.com]

I could, but I'd have to go to Unicode.org to do so.

[pne.livejournal.com]

You think?

I thought unicode.org only had things such as "BRAILLE PATTERN DOTS-124" as character names, which isn't very helpful.

*goes and gets The Book*

Hm, this says "There is no fixed correspondence between a dot pattern and a character or symbol of any given script. Dot pattern assignments are dependent on context and user community." ... "The assignment of meanings to Braille patterns is outside the scope of this standard".

Still, if you manage to read the text by information on Unicode.org, I'd be interested in how you managed.

(I've used a similar approach with alphabetic scripts, though, by pasting the text into UniPad and advancing the cursor one character at a time and looking at the Unicode character name in the status bar each time in an attempt to decipher the text.)

[darth-spacey.livejournal.com]

Well, dang. I was basing my claim on knowledge that Unicode contained Braille. I was also under the impression that the Braille writing system *did* include semantic information. I mean, M Braille, devising his alphabet, assigned letters to dot patterns. It's not his fault later people used his printing method for other reasons.

I'm not sure whether I understand the Consortium's stance on this. I mean, based on that, why doesn't Unicode encode a single circular glyph, and say "used in the Latin and Cyrillic Alphabets, Myanmar, and several other systems. Also used as a symbol for various reasons. The assignment of meanings to graphic shapes is outside the scope of this standard"?

[pne.livejournal.com]

I presume that reasons include (a) backwards compatibility with existing coded character sets and (b) the fact that it's not possible to systematically enumerate all possible alphabet glyphs the way it is with Braille patterns.

Another reason may be that Braille is used for things such as music, chemistry, mathematics, and other non-linguistic notation systems, so saying that a particular pattern represents, say, a "Q" doesn't make that much sense there.

What does work is if you know how M. Braille (dot after M since the last letter of the word is not part of the abbreviation) assigned letters to patterns, this will help you to some extent with other languages since most languages tend to use, say, dots-145 ("D" in French Braille) for some D-like letter (e.g. dal in Arabic, deh in Cyrillic, delta in Greek, etc.).

[pne.livejournal.com]

I was also under the impression that the Braille writing system *did* include semantic information.

Part of the problem may be that, due to the many uses to which Braille has been put, any given pattern has multiple pieces of semantic information attached to it, depending on the context in which it is used.

Sort of like annotating U+0043 "C" with its meaning as a school grade, abbreviation for "see" in some varieties of informal English, letter of the Latin alphabet, musical note, etc. etc.

[pthalo]

and in braille, dots 14 could be "can" "c" the number 3, depending on what precedes it, "cannot" if 456 is before it... and something else entirely in German braille, right? since "c" has to be escaped?

[pne.livejournal.com]

and something else entirely in German braille, right? since "c" has to be escaped?

Yup. It's "en" in the German equivalent of Grade 2 braille.

[pthalo]

that makes sense. how are umlauts handled?

umlauts in German Braille

[pne.livejournal.com]

Umlauts are separate letters, and are included even in grade 1 braille.

ä is dots-345 ("ar" in English); ö is dots-246 ("ow" in English); ü is dots-1256 ("ou" in English); and ß is dots-2346 ("the" in English).

In German grade 2 braille after the spelling reform, though, dots-2346 stands for "ss" since "ß" is less common now than it was previously, so if you want "ß" you have to escape it with dot-6.

[allegrox.livejournal.com]

Try Omniglot (http://www.omniglot.com/writing/braille.htm). That's the best source I've found for deciphering scripts I don't know.

[pthalo]

Kynn was here

[pne.livejournal.com]

Yay! u r smrt!

[pthalo]

but why would they make a Capatcha in Braille?

[pne.livejournal.com]

As an example of how inaccessible captchas can be, I suppose. (For example, to people who don't speak a language well or have visual impairments.)

I very much doubt that the image was of a serious captcha. The entire presentation was about how captchas have problems.

[pthalo]

ah! that makes more sense then. i did read through the presentation but it still sort of made me boggle.

[allegrox.livejournal.com]

Of course, I've found that Captchas don't have to be in an unfamiliar script to be unintelligible. They're sometimes just garbled that badly. I once had to use the audio alternative, even though I have 20-20 vision.

[codeman38.livejournal.com]

kynn used the Braille captcha on his (now apparently defunct) Maccessibility blog precisely to make a point of how inaccessible captchas are. Of course, it was quite easy to decode— on the original blog, the ALT attribute of the Braille image was "Kynn was here"!

[pne.livejournal.com]

Ah, so that's where the author of the presentation got it from. (The captcha does say "Kynn was here" and I was wondering who Kynn was.)