Language Weaver in the World of Google Translation
January 6, 2010
Several years ago, I took a look at Language Weaver, founded in 2002 by some wizards from the Univ3rsity of Southern California. The company’s technology struck me as more up to date than Systran, the engine that powers Yahoo’s Babelfish. In-Q-Tel pumped some money into the firm as well. Then Language Weaver caught my attention with its 2008 estimate that the machine translation market would tap into a $67 billion market. I remember watching an interview with Mark Tapling on Fox. The video is still available here. There were not too many details, but the number not long after the start of the financial meltdown in 2008 was an eye catcher.
The firm’s positioning is:
Language Weaver provides trusted automated translation solutions for high-value, dynamic digital information to improve human communications. Delivering a trusted level of translation quality, Language Weaver ensures that organizations maintain and extend brand voice across global media types and audiences.
The company now has a Wikipedia entry which strikes me as quite useful. You can read it here, and I won’t recycle that information. The company offers translation for about two dozen languages, and it offers what I think is an interesting software method that aligns translated documents at the segment or chunk level. This works, in my opinion, a bit like a knowledge base with some semi autonomous functions improving the core translation system.
The system can handle rich media; that is, “listening” to a podcast and translating the content. Language Weaver uses algorithmic methods. As computing horsepower creeps up, the system improves. The underlying method is statistical, so fast computers yield better translations.
What caught my attention on a routine updating of my search and content processing files was the fact that a number of links on the company’s home page did not work.
I was able to access an article by CEO Mark Tapling called “Building Loyalty after the Sale with Customer Driven Support Channels and Languages.” and “Dissolving Customer Support Communication Barriers.”
Like other search and content processing firms, the one-size-fits-all solution seems to be expressed in terms of solutions that solve specific problems; for example, voice of the customer or customer support problems.
Kirk and a Klingon need to avoid a failure to communicate. Image source: http://farm4.static.flickr.com/3380/3533169483_bae603ca18.jpg
When I reviewed various links on the company’s Web site, I located a list of the firm’s partners. This was interesting to me and included a number of firms whose marketing people had neglected to tell me that their employers’ translation functionality was licensed from Language Weaver. You can find this list on the Language Weaver current partners page.
Several surprises for me were the inclusion of Attensity, BBN, Right Now, SPSS, and ZyLAB. Other licensees include:
- Clay Tablet, a translation connector that hooks translation systems together
- SDL, a global information management solutions company
- WorldLingo, a company licensing its technology to Microsoft
The company has a Web log, it features a post every week. You can access the blog here.
My view is that Language Weaver has a method that struck me as having some similarities to the one I explored in my Google monographs. The question becomes, “Can a company like Language Weaver continue to thrive when Google offers increasingly robust translation services without charge?”
I don’t know the answer to this question, but Language Weaver may become an outfit that offers an alternative to Google. Microsoft is moving forward with translation functionality . Will another outfit step up and buy Language Weaver? My hunch is that there will be some companies willing to pay for a non Google solution for now.
As Google’s translation system gains traction, will Language Weaver’s value ratchet upwards, or will it like other translation services hit a glass ceiling? I don’t know the answer. More information when I have it.
Stephen E. Arnold, January 4, 2010
A freebie. I will report this to the US Department of State. That outfit tracks translation systems once in a while.
Comments
8 Responses to “Language Weaver in the World of Google Translation”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
Your blog article looks suspiciously like a PR piece disguised as a real blog article. I apologize upfront if this is not the case.
I think you may not be aware that SMT technology is available in open source (search for Moses and SMT) and you may not be aware that now there are many new companies that appear to have much more compelling or at least more cost-effective offerings than Language Weaver. Thus, your hunch that people will pay for non-Google solutions is quite right, there are at least 5 such companies with SMT technology.
I am a former VP of Sales & Marketing of Language Weaver who has moved on to a new and exciting company called Asia Online that provides a much more comprehensive “platform” that enables many more people to build and constantly tune SMT engines in a way that has never been possible before to quality levels that have never been seen before.
Here is a link that describes this emerging translation platform http://is.gd/5MjDe written by a leading industry analyst. This is a platform that is designed to be used by both language industry professionals and can be embedded into CRM and global customer facing applications to facilitate global communications.
Google may be an excellent option for a casual user who wants an occasional web page or email translated but it is far from ready as an enterprise ready application. Also, there is no possibility of using Google if you have any concern about data privacy and they are not likely to tune it for individual corporate customers for very specific customer needs. Google is a one-size-fits-all approach but I think it is likely that they already produce the best baselines. You can easily see this by comparing Google translations to most others available for free on the web.
Also given that the current ACTUAL market size for machine translation software is around $50 Million in 2008 – the estimate of $67Billion is somewhat silly. There is an extensive discussion of this in a Linked In MT discussion group that you can find here: http://bit.ly/MTmktSize and here http://bit.ly/MTFuture
Thank you for making people aware of the possibilities of MT in Enterprise Use. I suspect that it would be more useful if you provided a more complete picture and I would suggest that you might find the information in the Linked In Automated Language Translation Group ( http://is.gd/5Peuf) quite useful as multiple vendors are present and all speaking and sharing information in a much more comprehensive and complete way to add you to your initial summary.
Thanks
This alignment-based translation is the same kind of thing that was invented at IBM back in the 1990s. It’s been genealized by the folks behind LanguageWeaver and others to incorporate phrases/chunks as well as single words.
For a bit of history, check out:
http://en.wikipedia.org/wiki/Statistical_machine_translation
Google itself has been working on similar research with good results:
http://research.google.com/pubs/och.html
Besides Language Weaver, there are other companies such as Asia Online (www.asiaonline.net) or us, Ta with you (www.tauyou.com), who have very interesting solutions as well.
Our approach is unique because we build completely customized solutions for our clients, in any language and with limited data. We are specialists in hybrid (statistical + rule-based) machine translation technology, offering confidentiality and complete control of the data that feeds the system of any client (a valuable point with respect to Google), among other key features.
With as little as 1 million translated words from the client, we have developed
+70% accuracy solutions, but this depends on the domain and the quality of the data, among others.
Kirti Vashee,
Ah, public relations. I wish I could agree that my write was PR, but I crank out these articles based on references I find in other documents. You may get a better sense of what I am doing if you read the About section of the blog. PR people get paid. You may notice the disclaimers at the foot of each article which make clear that I labor without compensation. If this eludes you, may I suggest you stop reading the Web log?
Stephen E Arnold
January 7, 2010
Stephen
Again, I apologize for suggesting that your post was anything but a summary of your own thoughts.
I think that you and your readers may find some of the following links as a way to expand on the great start that you have given them in your initial summary. I agree that the technology has great promise (not $67B worth) but clearly great promise and potential.
The market for “highly customized” MT systems is just beginning and Language Weaver is not a leader. I hope that Asia Online builds a leadership position here and I humbly request that you keep your eye on us.
http://www.languagestudio.com is where much of the information updates will happen related to the Asia Online platform.
Also perhaps interesting are some video presentations involving MT from a recent conference that talk about the promise of highly interactive man-machine collaborations. Enterprise quality will require much more man-machine collaboration than Google allows and we believe that really compelling translation quality can only come from much more intensive, structured and managed human feedback. This will allow global enterprises to expand the global dialogue and share information across language on a scale not seen before.
Implementing large scale Machine Translation in Patent Information by Andrew Ruffner CTO Lexis Nexis on comparison of Japanese Patent MT systems http://dotsub.com/view/159ce97c-dbd4-4d6a-90c2-427a3a3e755f
The On-going Evolution of the Localization Business by Renato Beninatto CEO Milengo
http://dotsub.com/view/7da1f3a0-4df2-45a4-b62d-99434c2cf75f
Beyond Machine Translation: Collaboration, Integration, Quality, Change and Jobs – CEO Dion Wiggins on Asia Online Vision
http://dotsub.com/view/727cdacf-9653-40ef-b6f3-6145ca107db0
I am sure that Diego and others could add more information to show that the whole segment of producing “high quality customized and tuned MT “is growing and that the market is doing well possibly even because of the visibility that Google and Microsoft are giving to MT.
Stephen
Again, I apologize for suggesting that your post was anything but a summary of your own thoughts.
I think that you and your readers may find some of the following links interesting as a way to expand on the great start that you have given them in your initial summary. I agree that the technology has great promise (not $67B worth) but clearly great promise and potential.
The market for “highly customized” MT systems is just beginning and Language Weaver is not a leader. I hope that Asia Online builds a leadership position here and I humbly request that you keep your eye on us.
http://www.languagestudio.com is where much of the information updates will happen related to the Asia Online platform.
Also perhaps interesting are some video presentations involving MT from a recent conference that talk about the promise of highly interactive man-machine collaborations. Enterprise quality will require much more man-machine collaboration than Google allows and we believe that really compelling translation quality can only come from much more intensive, structured and managed human feedback. This will allow global enterprises to expand the global dialogue and share information across language on a scale not seen before.
Implementing large scale Machine Translation in Patent Information by Andrew Ruffner CTO Lexis Nexis on comparison of Japanese Patent MT systems http://dotsub.com/view/159ce97c-dbd4-4d6a-90c2-427a3a3e755f
The On-going Evolution of the Localization Business by Renato Beninatto CEO Milengo
http://dotsub.com/view/7da1f3a0-4df2-45a4-b62d-99434c2cf75f
Beyond Machine Translation: Collaboration, Integration, Quality, Change and Jobs – CEO Dion Wiggins on Asia Online Vision
http://dotsub.com/view/727cdacf-9653-40ef-b6f3-6145ca107db0
I am sure that Diego and others could add more information to show that the whole segment of producing “high quality customized and tuned MT “is growing and that the market is doing well possibly even because of the visibility that Google and Microsoft are giving to MT.
Kirti Vashee,
You are entitled to your opinions. My grandmother told me, “Never complain, never explain.” You did both.
What would grandma think? I don’t do PR. I don’t do much. At age 65, I am lucky to write blog posts and respond to the sharp wit and telling comments. I leave PR to the 20 somethings, the azure chip consultants, and the pre Wal*Mart greeters.
Stephen Arnold, January 7, 2010
@Stephen
Why are you being so defensive? Seems like you’re attacking Kirti for simply expressing an opinion.
What wasn’t clear from reading the article was the answer to the question, “Is there a future for translation businesses in a world of Google Translation?” (at least not imo).