Friday 25 January 2008

[ F K ] Interesting Machine Translation Project by a Karachite

Interesting Machine Translation Project by a Karachite

Zeeshan Ahmed is a young engineer who is actively working on an interesting machine translation project called PakTranslations. Self described as a die hard Karachiwala, Zeeshan is a graudate engineer from GIKI

KMB has conducted an email interview with Zeeshan on this project and what he expects out of it in the coming days. In a time when we have a tough fight between the forces of gloom and that of hope, let us do every bit possible to side with the positive forces.

Interview with PakTranslations follows.


KMB: What is Pak Translations?

Zeeshan: Pak
Translations.com is a personal project. Something I set out to do a year an a half ago. Personally developed the technology and got the machine translation dictionary done by volunteers, friends and data-entry people across Karachi sitting in photo-stat shops. With the new version launched in a few days, then my prime focus would be on bringing in traffic and introducing Sindhi.

For more info, see the mission here and the road map here.

KMB: Tells us something more about yourself that is included in you online profile?

Zeeshan: Well I am a Computer Systems Engineer, graduated in 2004. Did a few jobs first, I call it 'paid my dues'. It was in May 2006 when I realized that one should save his best for his own work. No one else deserves it and would never appreciate it. Therefore I decided to dedicated a portion of my career on work, rather than a job. It has been tough but over all a great experience!

KMB: Is Pak Translations just a word by word replacement of English words with Urdu words? or something more than that?

Zeeshan: Its not word to word. It rearranges, supports tenses, grammar, gender, respect and other language details. The technology is much advance than it looks. Its actually the data which lacks. That would improve with time and with the new version coming, the system would also become context aware.

The translation core uses rule-based grammar. This was the only way I could do it. As Urdu has not been much worked on, there is not much digital data available. The statistical approach (Google's way) needs millions of examples to train the system. Therefore I went for the age old rule based approach for translation and statistical for learning. Learning would kick in with the new version.

KMB: What is the nature of the project? (Commercial, hobby, passion or a cross between these)

Zeeshan: Started as a passion but yes now it is a cross between everything. Still, commercial or no commercial, technology would always be worked on and the end result would always be free for the end user.

KMB: Do you agree that once machine translation for English/Urdu is perfected to the reasonable level, its main pull with come from mobile population?

Zeeshan: Being a very new concept for Pakistan, I actually didn't know where the real pull would come from. And I didn't spend much time trying to figure it out. I would keep improving the technology and see where the light bulbs go off. Can be in media, news papers, dot com, ISPs, etc. But I do agree that mobile market is one where I have high hopes from.

KMB: How does it feel to have Pak Translations at the first place in the Google's Zeitgeist this month?

Zeeshan: haha ... may sound indifferent of me but I didnt know about Google Zeitgeist till I got your mail. Then I Googled and found out. This of course is an honor but actually makes my mind wonder of the possibilities ahead. i.e. when the technology is matured and proper marketing done.

KMB: Do you have a dream of someone from GMY acquiring Pak Translations in future?

Zeeshan: As you may have guessed already that I am not much of a dot com person. About acquisition, of course but only if

- the technology is constantly worked on

- free for end user

- price is right

 

 
 
 
Fun_Karachi Yahoo Group [61,000+ Members Worldwide]
Sharing Fun, Love, Information & Quality stuff by Email

 
 
 
---- > To start receiving our emails, please send a blank email at fun_karachi-subscribe@yahoogroups.com
 

__._,_.___

Your email settings: Individual Email|Traditional
Change settings via the Web (Yahoo! ID required)
Change settings via email: Switch delivery to Daily Digest | Switch to Fully Featured
Visit Your Group | Yahoo! Groups Terms of Use | Unsubscribe

__,_._,___

No comments: