As I metnioned at yesterday's MELCOM meting, there may be some interest in 
the following message from Google - see also the comments below by a couple 
of US-based librarians. I have performed a couple of tests, and Yamli is 
definitely more sophisticated and accurate than Google Transliteration, 
although Google offers more languages and scripts. Fore example, Google 
could not transliterate the Arabic word shara'i` (plural of shari`ah) accurately, 
while Yamli offered several Arabic script alternatives to the English word, 
including the correct one.
 

Paul Auchterlonie.

Librarian for Midde East Studies,
University of Exeter.




Blog: Official Google Blog 
Post: Link: http://googleblog.blogspot.com/2009/12/transliteration-goes-
global.html 



Transliteration goes global 

12/17/2009 10:34:00 AM 

Most of us use a keyboard to enter text; it's one of the most basic activities 
we perform on a computer. However even this simple activity can be 
cumbersome in many parts of the world. If you've ever tried to type in a non-
Roman script using a Roman keyboard, you know that it can be difficult to do. 
Many of us at Google's Bangalore office experienced this problem firsthand. 
Roman keyboards are the norm in India, making it difficult to type in Indian 
languages. We decided to tackle this problem by making it very easy to type 
phonetically using Roman characters and we launched this service as Google 
Transliteration.


Using Google Transliteration you can convert Roman characters to their 
phonetic equivalent in your language. Note that this is not the same as 
translation — it's the sound of the words that are converted from one 
alphabet to the other. For example, typing "hamesha" transliterates into Hindi 
as: , typing "salaam" transliterates into Persian as: and typing "spasibo" 
transliterates into Russian as . Since our initial launch for a single Indian 
language, we've been hard at work on improving quality, adding more 
languages and new features.

Today we are pleased to introduce a new and improved version of Google 
Transliteration, available in Google Labs or at 
http://www.google.com/transliterate. 

 

In this new version, you can select from one of seventeen supported 
languages: Arabic, Bengali, Greek, Gujarati, Hindi, Kannada, Malayalam, 
Marathi, Nepali, Persian, Punjabi, Russian, Sanskrit, Serbian, Tamil, Telugu and 
Urdu. You can also compose richly formatted text and look up word definitions 
with our dictionary integration. If the default transliteration is not the word 
you wanted, you can highlight it to see a list of alternatives. For even finer-
grained control, we provide a unicode character picker to allow character-by-
character composition. 

 

Google Transliteration is integrated into several Google properties and we have 
an API and bookmarklets to extend this capability to other websites. A 
solution we initially built to solve a problem we saw here in India is now being 
used in many other parts of the world as well - one small example of the scale 
and leverage that technology can bring in today's increasingly globalized 
environment. As with all labs products, we will continue to improve the 
technology and try out new features. We would love to hear from you, so do 
let us know what you think.


Posted by Nilesh Tathawadekar and Mohammed Aslam, Software Engineers

 

 

Mark Muhlhausler wrote : 

 

A similar feature, paired with an Arabic search engine, has been around 
for a while:
http://www.yamli.com/
... functions remarkably well.



Andras Riedlmayer wrote:
Just tried a quick test of Google Transliterate on some of the languages. 
Arabic worked remarkably well, even based on non-standard romanization. 
Russian and Serbian (Cyrillis script) also seems to work.

But the Persian version had problems with some words of Arabic origin. 
When I tried typing in the name of the nineteenth-century Iranian ruler 
Muzaffar al-Din Shah, it did not produce correct Arabic script no matter 
what form of romanization I tried (muzaffaruddin, mozaffaroddin, mozaffar 
al-din, etc.). The Arabic Google transliterator had no trouble reproducing 
the word or the construct correctly. I hope this is something they'll 
fix soon.

On the other hand, both the Arabic and the Persian transliterators were 
good at producing the correct spellings of words with the various letters 
that can be pronounced as "S" or "Z" in the standard or the vernacular, 
(also picks the right spelling for words with initial GH vs Q in Persian).


 


--