I have an idea. Auto correct feature of Avro 3 is a very nice. Can it not be extended to produce a translator for the Engling language?

I mean I will write 'My name is Debashis' and Avro, using its auto-correct feature, will write the equivalent Bangla 'amar nam debashis'.

Actually, I have tried it with current Avro distribution and it worked nicely :-)

The problem is with current auto-correct feature you can not insert an item without a replacement, so the written Bangla becomes something like: amar nam holo debashis.

But we don't use 'holo' in place of 'is' when we speak normaly in Bangla!

Another problem is that there can be more than one meaning of an English word in Bangla.

To overcome these problems I have some suggestions:

1. Adjust the auto-correct feature to include an item with an empty replacement.
2. Extend it to have more than one replacement for a single entry.
3. Typing the entry with more than one replacement will show up a list /combo box with the replacements from where we can select what we want - this will be a natural extension of the preview window.
4. Somewhere in the forum I saw a link of a English-Bangla dictionary. If they are willing to provide Avro with the mapping of English to Bangla words, then it will be very nice of them. We are not asking the meaning after all.

There are some problems though:

1. Different tense of English will effect the bangla verbs itslef. For example:
I did this: ami eTa korechhi
I will do this: ami eTa kor`bo.
Do note the change in the Bangla verb 'korechhi' and korbo'. Probably it can be over come by using the multiple entry: one for 'did' -> 'korechhi' and another for 'will do' -> kor`bo'. But you have to change the auto-correct feature. Probably, your current .dct file does not allow replacing two words at a single go.

2. May have to provide two keyboard layouts: one for 'kathya' and another for 'sadhu' Bangla bhasha. Keeping in mind that the verbs will also change accordingly. I don't know at what extent it will be possible for you to implement this with your current architecture. And not to mention that there are so many language different between different region within Bangla speaking people.

3. Not very easy to translate to Bangla 'bhabartha'. But, you can provide a translation scheme from English to Bangla and which is not a simple work anyway. But what Avro did was not so simple either. So can we expect that Avro will think on this topic?

Thank you Debashis, I think auto-correct is not actually the perfect way for translation. We may have to start a new project for the translations work from English to Bangla.

The fact is, i've not found any research work regarding this. So we need a good number of manpower who can help us voluntarily to find different sentence structures, to find out grammatical error etc.(ofcourse, this project will be freeware).

The first problem I found while thinking about your suggestion-
What will be the Bangla of "You"? -"apni", "tumi" or "tui"? Situation may occure where a single sentance can mean "tumi" and "tui" at different places.

Any idea?

Yes, that's a problem.

That is why I suggested the combo box with all thee replacement. Actually this problem is less difficult than the problem with verbs. Verbs has tense associated with them and there are so many tenses....

Then there is the actual word mapping problem. If you can't get the ready, someone have to actually create the mapping by typing around 30,000 English to Bangla word mapping! That's not easy at all.

Since, the ground work, the auto-correct feature, is ready for the translation, hope you will think on this.

And man-power is the real problem.

I think you can help

We have already taken a dictionary project for Avro Keyboard. In any future release, you will get a Bangla spell checker... And we may also extend the dictionary to make Avro Phonetic more flexible than present (e.g. Typing "akkhar" will automatically suggest for the correct word "okkhor"...). So I think, at present the dictionary project is not a big deal...

Let us find out the different sentence structures first for Bangla & English, someday we will manage to overcome the barriers which seems to be quite tough right now...

Let us know if you can help, I can invite other gurus then to take part...

Another thing, A combo list with all possible suggestions is not a good idea I think, the whole process must be automatic. Suppose we want to extend the translation as a web service (just like google.com does), then where would we place the suggestion combo for users to interact?

No problem, I will do whatever I can. Please, let me know if there is something that I can do. That will be my pleasure.

And for your dictionary project, I think the dictionary should/must be in place before Avro starts doing the translation keyboard map.

As far as the combo box goes, you are correct that it should be automated, but I was just thinking from an user point of view. May be we can think of some other ways.

Well, if I get some other ideas, I will let you know. Many thanks.

At first, we need to bring out the difference between English & Bangla sentence structures (basic sentences ofcourse, for the first time).

Any idea or suggestion or example?

I will think of it and let you know.

Thanks.

আমি বেশ আগে এধরণের একটা ওপেনসোর্স প্রজেক্ট দেখেছিলাম এবং সেইটার সোর্সকোডও আমি ডাউনলোড করেছিলাম। কিন্তু, এই মুহূর্তে কোথায় আছে আমাকে খুঁজতে হবে। পেলে সাবমিট করব। আমার মনে হয় সেটি পেলে আপনাদের কাজে লাগবে।

আর অনুবাদের ক্ষেত্রে আমার মতামত হলো, এটি যেহেতু একটি বড় প্রজেক্ট এবং খুব জরুরী নয়(স্পেল চেকার, ওসিআর এসবের তুলনায়)। তাই আমাদের উচিৎ আগে জরুরী বিষয়গুলো নিয়ে ভাবা।


রিফাত।

Wel, long ago I'd done some research about the English to Bangla translation. And from professaional view I realize & decide (As a CEO of Bangaliana) that "This project will need a lot's of time and work, if 10 person works with, they will need at least 1-2 years to build a good tranlation engine!"

Anyway, Though it's a big deal, and if there are a lot voluntier then certainly we make it up!


Some complex thingy we have to consider: "play" means "Khela" and "Bajano", which one should choosen? Depend on the sentence, so one have to work with words, sentences, verb, object, subject and lots more....

Though at primary lavel it's too easy to build-up....but in advance it's too hard to implement....I've some source code, sample project and some idea....

But like Bangla Bhai, I think we should spent our time with spell checker, OCR, Bangla Text to Voice engine....otherword....as the translation project needs more volunteer [typist] than programmer...it's could be start today...let OCR, Spell checker, Bangla TTS on the professional group/company, what you say ?

leave your comments....

যাক ভাই, তাহলে শেষ পর্যন্ত একমত হয়েছেন.....

Absolutely right. Translation needs more time and resources.

Well it can be done later, after the spell checker, OCR etc. Actually it depends on these.

It just came to my mind and I saw no one posted such a request, thus I have made a request.

Anyway, thanks for all your replies and times. And many thanks for such a good Bangla Typing software 'Avro'. I just love to have the spell-checker with it.