Wednesday, July 24, 2013

Jibbigo: Working Out the Arab Dialect


In order to help the US soldiers on ground for war and to understand the Arabic language, the US government came up with the idea of creating translation electronic software that would make it easy for US soldiers to communicate with the local populace there in a better way.

The task was a hard one. The researchers at Carnegie Mellon University were contacted to develop a system that would have to deal with a different Arabic dialect. The problems of different word choices, phonology and pronunciation by different people who speak Arabic have created many different dialects to such an extent that some would look at the Arabic language as a cluster of languages.
This progresses to the existing voice-recognition machines based on MSA, which is widely used by schools, academia and media.

Jibbigo is an app developed by Dr. Waibel’s Team to push down language barriers. The app has been fed with 40,000 words of vocabulary for voice to voice translation in English and voice to text translations in different languages.

This arduous task of developing such database is being completed by crowd sourcing online and on the ground. The vast resource of crowd sourcing is by the participation of Arabs, since the Arab spring, who now takes more part in blogging and social media platform, providing an ample pool of written and spoken dialects. Twitter and Microsoft are also on the roll by placing thousands of volunteers to the task of translating and localizing tweets and developing translation tools, respectively.

Researchers are keen to provide the Arabs with better tools to communicate because they need to voice their opinions, but don’t have an appropriate medium to do so.

One of the major problem arises with diacritics, as it is responsible for distinguishing among different Arabic dialects; it is often missing from the formal written Arabic. To observe patterns of words, modulations, diacritics and accents and to resolve contexts, the researchers need to build such modules that bring all the factors of diacritics and dialects into account.

Google translators have categorized Arabic into four major dialects comprising of: Egyptian, North African, Arab Gulf and Levantine (Syrian, Jordanian and Lebanese dialects). They, together with researchers at the Carnegie Mellon, have concluded that a single speech recognition system is not flexible for all the dialects as they are completely acoustically different.

Researchers believe the dialects of various regions are distinct to such an extent that a Moroccan cannot understand someone from the Gulf. If we recognize those differences in phonetic habit, we can easily identify the speaker’s regional dialect even when they speak Modern Standard Arabic, Classical Arabic or English.
For example, An Egyptian accents every word in a way that even if he is speaking English, a person can easily judge them that he is from Egypt.

It is an important tool to have Google’s module for Gulf Arabic as it contains about 60,000 regularly-used English words. Developers are forced to model the variations that Arabic speakers gain freely in their dialects and there lies the main problem.


Researchers view the development and perfection of Dialectical databases as a long term project. It is especially for the people in remote areas where there is no published form of the language they speak, and without this effort their dialect would die out. If that happens, unfortunately the Arab world won’t be able to communicate with each other through technology. 
--
TextpoweR Creative Arabic Translation & Copywriting Services has a professional team of experienced Arabic Translators and Arabic Copywriters, who can provide you with high quality Arabic translations and copies of your legal, technical, financial, medical, marketing and other documents and materials. The team is based in Dubai, United Arab Emirates (UAE), but serving several happy clients around the world from Europe, USA, Middle East and Asia.

1 comment:

  1. The world's most popular translation software is free... By this soft you can get -
    1 click translation
    77 languages
    1,700 free dictionaries
    Wikipedia at a click
    Get It Now

    ReplyDelete