Watch videos with subtitles in your language, upload your videos, create your own subtitles! Click here to learn more on "how to Dotsub"

Beyond Machine Translation: Collaboration, Integration, Quality, Change and Jobs

0 (0 Likes / 0 Dislikes)
Good morning everyone, thank you for this opportunity to present today. I am going to present today on a number of different angles and demonstrate some of the things that have been going on in the Asia Online world that we are now delivering to our customers I see from Renato's [Renato Beninatto] opinion we already have a new fan which is great but we are starting to make these tools now readily available. Nelson Mandela - he once said when you speak to a man in a language he understands it goes to his head. But when you speak to him in his own language that goes to his heart! That's a really important statement Many of the people in this room are bilingual, trilingual or many more But the language you understand the most and the culture you understand the most is your own. You always get the most information out of something in your own language Machine translation has been around for a very long time It's ironic the abbreviation is MT. Machine translation has actually been around for about 50 years. But it has been 50 years of empty promises So it's quite appropriate, today things are changing. Google has made machine translation acceptable in the eyes of many people. We like Google, it's our marketing department. It shows people what machine translation is capable of. But if you want high quality you need to go to the next step. So there is an industry that has been around and failing for 50 years. You have to ask why does it still exist? What other industry can last for 50 years by constantly failing? It is very simple. There is an infinite demand, it has never gone away. In fact it keeps growing The demand for translation is growing by the day. Whats been missing is quality. But machine translation without humans is a real problem. On the left it actually says "restaurant" It is a good example. The Chinese people who translated this billboard, obviously didn't do any QA so very simple Translation Server Error And even that is translated poorly, it says "translate server error" So there's been an evolution in machine translation Rules based technology continues to slowly get better and in recent years they have been combining it with statistical approaches and getting some more advantages 9/11 happened and a big push changed with machine translation Many governments said we now need to know, with a high throughput very quickly what other people are saying in other languages and we have huge volumes of data. Several people took on that challenge The first real statitical phase based commercial approach was undertaken and that was good for it's time. Then they started looking at how do we solve some of the additional challenges and things like syntax came in We brought something new to the game. It's called "clean data statistical machine translation" and its a very simple principal. If you have a limited volume of data but it's better and cleaner than a bucket full of garbage, the chances to have an error are fewer. Why would software built for translation be different from any other software where the golden rule is: "Garbage in - garbage out" This is a data problem But it's a quality problem too of the data that you are feeding it. Then we added other techniques such as syntax, grammatical reordering and other techniques that are now delivering very very high quality. The market is worth about 16.7 billion US dollars. And is forecast to grow to about 24 billion by 2012. That's the human translation services industry. Machine translation is forecast to be just 95 (million US$) It is very low. Why? The demand is there. Until 2009, quality was missing. There are several companies now that are reaching a quality threshold where machine translation can be useful. Google is a good example. Previously they used the Systran engine that many of you may have seen as Babelfish. That is still up there today. But the quality is not there in many languages. Then statistics started coming in. Google is gathering more and more data. Great!! It's making machine translation available to people with what we call gist based translations. And it continues to improve. That quality threshold is important. It is a market inflection point. I believe that this forecast for machine translation is not quite right. Once you achieve a quality result you also start to achieve greater sales. And if the cost is right you can do something like what Andrew from LexisNexis talked about yesterday. Translating 120,000 documents a day that he can then go and sell and monetize. There is many industries where machine translation can start to take hold. We are getting a lot of interest in the travel industry. Technical support with knowledge base, where its not just translated because it needs to be there it is translated to save money or generate money. In the travel industry if you have hotel reviews or descriptions online in foreign languages. More people come to your site. More people buy tickets. If you are doing technical support, if you make your knowledge available, lets say in Thai. You no longer need to pick up the phone to call Microsoft or Cisco or whoever you are calling. That cuts their costs. So this space is moving very very quickly. This one slide really sums it up. Up the top of the pyramid is what humans are capable of doing today. two to three thousand words per day. And often parts of products only. Microsoft Office in Thai is a great example. It is wonderful, you are tapping away in Excel all the user interface is in Thai. Then you get stuck on a formula. You press F1... oh no - English. All the help Why? Because translating all of those help files into a tier below mainstream languages is very expensive. And not only that, it is very large and time consuming. Many of these things are partially translated.

Video Details

Duration: 32 minutes and 16 seconds
Year: 2009
Country: Thailand
Language: English
Producer: Asia Online
Views: 405
Posted by: asiaonline on Dec 11, 2009

Localization and Translation Thailand 2009
Day 2 Beyond Machine Translation: Collaboration, Integration, Quality, Change and Jobs
- Dion Wiggins, CEO, Asia Online
Twitter: #LTBKK

Caption and Translate

    Sign In/Register for Dotsub above to caption this video.