Language Weaver Co-Founder and Chief Technology Officer Daniel Marcu says that rapid advancements in the field of software-based statistical translation are providing government customers with a range of new, more powerful translation solutions. 

Daniel is a recognized leader in natural language processing.

In the following podcast interview, Daniel says that both the speed and accuracy of statistical translation systems have improved exponentially in the past 10 years. He also explains how Language Weaver is working to further improve the applicability of its proprietary translation technology:

Subscribe to the Language Weaver Podcast.

Technorati Tags: Language Weaver, Statistical Machine Translation, Translation, Daniel Marcu

In Berlin at Localization World, a global conference that focuses on translation and localization, Language Weaver CEO Mark Tapling said today that the future is bright for Language Service Providers (LSPs) who embrace statistical machine translation as a tool to greatly expand business and increase margins.

CEO MARK TAPLING

“The Language Weaver value proposition consists of three key elements – speed, volume and accuracy,” said Tapling. “Language Service Providers (LSPs) share the same goals – and so we’re able to provide great value to our LSP partners.” 

“With our technology we continue to broaden the amount of content that is translated,” said Tapling. ”However, the human linguist will always have a high value-add role to play in the process.”

You can listen to an 8 minute interview with Mark Tapling here, or Subscribe to the Language Weaver Podcast.

Technorati Tags: Translations, translation, Language Weaver, Localization, Localization World, Language Service Providers, Mark Tapling

We introduced Kontrib.com early last year with the hopes of putting a new spin on a popular Internet concept known as social bookmarking. By introducing automatic language translation into the mix, Language Weaver created the world’s first multi-lingual social bookmarking website with the aim of breaking down the language barrier on the Internet. Suddenly individuals could share content in one of several languages and have that very same content be translated into multiple languages automatically - leveraging the original content several times over. Someone who posted a story in English would have that very same story be made available in Arabic, Chinese, French, and Spanish.

While the translations were sometimes far from perfect, readers were still coming to the website. Why you might ask? With Kontrib, users at least got some understanding or “gist” of what another person was trying to convey in a different language. That alone goes a long way towards breaking down the language barrier.

This of course also has deep implications for online web-sites and publications! By adding automated translation to your website, you are opening the doors to many more readers which in turn translates to higher traffic numbers and hopefully higher ad revenues.

What better way to show this than to go over some of the website statistics of Kontrib.com that has been collected over the last 6+ months.

Back in July of 2007, the top 10 countries coming to Kontrib were as follows:

1. United States
2. Canada
3. France
4. United Kingdom
5. Saudi Arabia
6. Egypt
7. Morocco
8. Mexico
9. Australia
10. Brazil

This list coincided nicely with the languages that were being offered by Kontrib which included at the time:

1. English
2. French
3. Spanish
4. Arabic

In late July, we added Romanian (Romanian -> English and English -> Romanian) and Chinese (Chinese -> English and English -> Chinese) to the mix and by October of 2007, we were getting the following rankings:

1. United States
2. Canada
3. Romania
4. Saudi Arabia
5. France
6. United Kingdom
7. Egypt
8. Mexico
9. Morocco
10. Spain

Interestingly enough, Romania started climbing up the list quite rapidly. But where was China you might ask? Well it turns out that Kontrib.com was actively being blocked by ISPs within China. Despite the blockage, China ranked 14th in the month of July and then 13th by October. More importantly, the # of visitors increased from 16 in July to 74 in September (still rather small of course but it was growing!)

Want more data? Sure thing!

In October we added bi-directional support for the following languages:

• Persian <-> English
• Hindi <-> English

During that same month, Iran ranked 51st on the list and India ranked 23rd.

Fast forward to the end of December, Iran was ranked 7th on our top ten list, and India was ranked 22 on our list. More importantly, the # of visitors from these countries jumped. Iran jumped from 11 visitors in Oct. to 159 visitors in December. Visitors from India jumped from 16 in October to 44 in December.

Not convinced? Here’s one last datapoint..

In November, we began offering bi-directional Italian <-> English support on Kontrib.com. When October ended, Italy ranked 28th overall on our list with 29 visitors. By the end of December, Italy ranked 15th on the list and we had received 70 visitors to the website!

To recap, back in August of 2007, we had the following top 20 list.

1. United States
2. Canada
3. Saudi Arabia
4. Mexico
5. France
6. United Kingdom
7. Egypt
8. Romania
9. Spain
10. Morocco
11. Turkey
12. United Arab Emirates
13. Brazil
14. China
15. Germany
16. Colombia
17. Dominican Republic
18. Australia
19. Puerto Rico
20. Netherlands

By the end of December of 2007, we had the following top 20 list.

1. United States
2. France 
3. Romania
4. Canada
5. Iran
6. Italy
7. United Kingdom
8. Saudi Arabia
9. Spain
10. Egypt
11. China 
12. Mexico
13. Brazil 
14. Germany
15. Morocco
16. Turkey
17. Belgium
18. Algeria
19. Australia
20. Sweden

We had five new countries to the top 20 list (which also pushed some other countries down the list) - yet more importantly, roughly only a quarter of the top 20 countries were English speaking - the rest were for a myriad of other languages.

This points to a couple of observations:

1. Automatic translations leverages your existing content and creates new content in different languages
2. Automatic translations help diversify website traffic - bringing in traffic that would otherwise not go to your website
3. Translated content will then be indexed by search engines which will also help drive traffic to your website
4. More readers from other countries drives up site traffic which COULD translate to higher revenues

As the language landscape of the Internet diversifies, so should your website and that’s where automatic translations can help make this happen.

No Tags

Over the past two years, Language Weaver has felt a definitive shift in the demand and perspectives on automated translation.  Despite our assertion that statistical machine translation (SMT) was a new methodology that would change the paradigm for machine translation, back in the fall of 2005, Language Weaver was still battling the previous negative perceptions that machine translation had built up as a result of 30 years of disappointing rule-based MT results.  While the government was seeing success with statistical MT, translators felt threatened, language service providers were skeptical at best, the commercial enterprise saw little value in what any type of MT could provide.

Fast forward to 2007…   

This fall, Language Weaver has attended several events around the world, and it is clear that the dark cloud over automated translation is starting to lift.  Companies approached several times in the past are now talking to us out of their own free will; translators are opening up to the idea of working with SMT; and Language Weaver has been invited to speak at many conferences and seminars.

Why the change?   

We see several factors that are contributing to the shift:

1. Global companies are feeling pressure to bring products to market faster, leading to greater demand for translation at lower costs
2. Microsoft has developed and successfully deployed its own version of statistically based translation software (SMT) to translate knowledge base content resulting in an increase in user satisfaction – truly a compelling argument to consider SMT!
3. Information is power – maybe a bit cliché, but exabytes of new digital data are being created annually and users want access to this information in their own language
4. More countries are coming “online” and the language of the internet continues to diversify. Users that speak languages other than English are beginning to dominate the web
5. Language Weaver has several successful deployments of customized SMT in translation productivity environments, leading to 150-400% increases in productivity.
6. Google keeps automated translation in front of web users with their own version of SMT.

Given these factors, it is clear that statistical MT can have a significant impact on global communications.  Language Weaver has always seen great promise in this technology, but now we think that the rest of the world is starting to share our view.  Stay tuned…2008 should be an interesting year!

No Tags