[accessibleimage] Re: Languages...

Hi Lisa,

The languages that ship with JAWS vary with the primary language of the
version that one buys.  If a person gets a copy in any of the languages that
Eloquence supports, US English, UK English, German, Italian, French,
Castilian Spanish, Latin American Spanish, Brazilian Portuguese, Finish and
Japanese, they will receive all of the others (Japanese is not shipped by
default as it is enormous, it isn't amazingly accurate and there isn't much
demand but Japanese customers get all of the rest).  Other languages like
Norwegian, Swedish, Arabic, Hangul, Thai, Tamil, Hindi, Icelandic, Canadian
French, Australian English, Dutch, Russian, Polish, Czech and others I
cannot remember depend upon the authors of the synthesizer included with
JAWS to support that language.  Almost all provide English but the variation
between one vendor and license agreement to another can be vast.  Arabic is
the most restrictive as the vendor charges nearly as much for the
synthesizer as one pays for JAWS and they only include one language.   

Some people buy JAWS and don't even get their native language.  Some of
these are speakers of very "popular" languages like Urdu, Swahili,
Cantonese, Mandarin, Hebrew, Ukrainian and others that I can't recall.
There are Braille tables for these languages so users can use JAWS with a
Braille display or in one of the languages that are available in a
synthesizer.

East Asian languages (other than Hangul) can be very difficult because of
the heuristics required to perform segmentation.  Japanese, for instance, is
a language without white space and little that a reader of a Indo-European
language would understand as punctuation.  A single phrase can contain all
four writing systems used in modern Japanese writing (Kanji, Katakana,
Hiragana, Romangi) and Kanji characters will change meaning if they are
alone or if they have one or more characters following them.  So, with no
spaces, a single character might be a word on its own and be followed by a
character which is a word on its own but, when next to each other, they are
a third word entirely.  A human discerns which word they are viewing by
context but a segmentation algorithm must do so by statistically analyzing
the document and determining the probability of it being word one, two or
three.  It obviously increases in difficulty as a third or fourth Kanji
character is added to the word.

Many years ago, in a different life entirely, I got to implement a
segmentation algorithm for a project I had been doing for the fort.  I am
not a computational linguist but enjoyed the task anyway.  It took about a
week to get running and about a year to refine and worked pretty well in a
number of different languages.  My system, however, could "chew on" an
entire document and hit the database of word frequency from the previous 50
years of newspapers from Tokyo, Osaka and Nagano as often as it liked and
for as long as it liked.  A JAWS user doesn't have the luxury of waiting for
a massively parallel computer to finish its analysis before speaking a
single word of an email.  So, at FS, we had to use a "high speed"
segmentation system which was fast but only about 65% accurate.

I've rambled again.  Someone should put me back in my cell.

Enjoy,
cdh
-----Original Message-----
From: accessibleimage-bounce@xxxxxxxxxxxxx
[mailto:accessibleimage-bounce@xxxxxxxxxxxxx] On Behalf Of Lisa Yayla
Sent: Thursday, January 12, 2006 10:06 AM
To: accessibleimage@xxxxxxxxxxxxx
Subject: [accessibleimage] Re: Languages...

Hi Chris,
Yes I'm pretty sure the list is constrained to plain text, but then I have
been wrong before. I thought text would keep it most accessible. I 'll
check it out. That's really neat about Jaws changing languages. I could
have edited the mail, but was lazy. Also very neat about the Klingon. How
many languages does Jaws come with? Can it swith from Italian to Norwegian
or does one have to install a new one for each language?
Regards,
Lisa

accessibleimage@xxxxxxxxxxxxx skriver:
>Lisa,
>
>Does this mailing list support html messages or we constrained to plain
>text?  While I can't get to this conference, I found it very interesting.
>Those of us who read using JAWS can have the synthesizer change language
>rules (from US English to Italian in my case reading this example if it
>was
>in html).  I'm not terribly multi-lingual but I can, in many European
>languages, glean the general meaning of text if it is pronounced
>relatively
>properly.  Thus, if you had sent an html message, we JAWS users would have
>the html language tag tell JAWS to switch from the English to the Italian
>synthesizer and back when it returned to English.
>
>This is a pretty cool feature which isn't too popular among Americans who
>think the world ends at our borders and that, even while traveling that
>people shouldn't talk foreign to them but our Canadian, European and Asian
>friends all seem to love it.  In this case, it isn't much of a real
>accessibility issue as the English was also provided but it is kind of fun
>when the synthesizer starts speaking with the proper accent and
>pronunciation rules.  Once, for fun, we worked with the guys at what was
>then Eloquent Technologies (it has been acquired by SpeechWorks which was
>later acquired by ScanSoft who, last year, changed their name to Nuance),
>to
>create a set of Klingon rules.  Our beta testers yelled that we were
>wasting
>time and that it was too silly to include in a very serious product like
>JAWS so we dropped it.  Some blinks have no sense of humor.
>
>Anyway, just a little tid bit of interest from the ancient land of
>uni-dimensional, one syllable at a time, G2 access technologies. 
>
>Have fun,
>cdh


Lisa Yayla
Huseby Kompetansesenter 
Oslo Norway
lisa.yayla@xxxxxxxxxx

Other related posts: