Thus Spaketh Idd Salim

In the quest for a definitive ANY-to-Swahili translator

by on Oct.22, 2010, under Coding, Personal

I had, have and always will be a big fan of Google. Everything they do is simple, well-thought, optimized, tested and done by the best of the best. So, naturally, I believe I am right beyond any possible reprieve to assume that ANYTHING that beats Google, is what we can call a real programming challenge.

I always test the Google-translate engine (English to Swahili) every now and then but the results are not all that impressive, even upto now.

Swahili. The hardest alphabet-based language to translate?


Swahili is a very structured and disciplined language and it’s translation defies the logic of common language translation algorithms. I can only ignorantly assume that Google have tried a combination of the best linguists they could get from YALE etc in Swahili + their endless resource of eubercoders to develop the engine. They are yet to hire real coders, who speak swahili. So, the logic flow between the linguists and coders gets polluted chinese-whisperesquely.

I have managed to source some support (moral, logical, conceptual and otherwise) from my boys, mentors and well-wishers and I think I can now officially start my quest to create a flawless ANY-to-SWAHILI translator. Text and voice.I will start the project in Dec, after a few Calculus remedial classes to polish up my maths.

I am yet to decide on the language, but I an inclined to use a combination of C++ and Java to build to the logic engine and weighting logic, then Python or plain old, cool PHP for the presentation layer. DB? MySQL of-course. Once done, web-services and APIs for Web, Wap, J2ME and android apps will follow.

I invite fellow coders who see this as a worthy challenge to join me in this task and let us put our beloved language out there and make it universal.

Habari ndiyo hiyo. Back to life.

Wazi.

:,

  • Fuyamode

    natural language processing noma, its right up there with machine learning. best case is word for word translation then “teach” the machine that your translation is better than its translation. lol punda yako ni mafuta sana hiyo ni noma ukiambia dame.

  • Ephantuk

    What about This? i will kick some ass – I kick baadhi ya punda

  • George

    The sad thing is that i have seen a Kenyan website use google translate for its page views in Swahili! It is ridiculous!!! Internationalization support can be a challenge and the semantic web is still a mile or 10 away for most African languages.

  • Anonymous

    No.. the sad thing is companies continuing to hire inexperienced campusers to do their sites and trusting them with their corporate identity.

  • Hellen

    “inexperienced campusers” very wrong term to use! In any case, practice makes perfect! Give the young people a chance

  • Anonymous

    I am sure what you mean is “Very humbling and ego-crushing term to use”.

    Like you said, they need practice, which takes time.. and time = experience…

    Hence “INEXPERIENCED”.

    Hapo vipi?

  • Simon

    it a good idea machine learning and natural language will be much needed so as to make well use of parallel text and transitive translations,thats involve word /sentence tagging from bilingual dictionary

  • http://joliea.wordpress.com/ Joliea

    ur tu-babies are cuuute :D

    but the google translate thing was just fooony!! hehe..

  • Anonymous

    @Joliea – Thanks.

    Punda yako ni mafuta sana…

Looking for something?

Use the form below to search the site:

Still not finding what you're looking for? Drop a comment on a post or contact us so we can take care of it!

Kenyan Blogs Webring Member

Possibly related