Lionbridge Partners with IBM for Real-Time Multilingual Solutions (Interview with Satish Maripuri, COO of Lionbridge)
On April 22nd Lionbridge and IBM announced their strategic agreement to accelerate the development and commercialization of a SaaS-based, Real-Time Translation Automation solution for the Commercial Markets.
We interviewed Lionbridge COO, Satish Maripuri to find out more about the announcement.
Satish - We are quite excited. With this technology, we will be adding a new offering that currently is not being serviced by the market so it’s being very well received. We are also seeing some pretty good interest in the new customer category that we had not anticipated.
CSN - Can you elaborate a little bit more on the new customer category you just mentioned?
Satish-Yes, perhaps the best way to answer that is if you take the translation needs in the market today and you customer stratification segment that market. Visualizing a pyramid, on the top of that pyramid you have what we call high-quality, just-in-case translation. Content at this segment has a very high cost of failure. Quality expectations are very high. The general trend there has been to translate that content with a fairly high consciousness to quality and of course as a result higher cost with human translation being the primary resource.
We have built our businesses, from a services standpoint, around human translation assisted with technology but still primarily utilizing human translation. However, this is still only a small portion of the overall content that exists within, and outside of, the enterprise. Industry estimates are showing that anywhere from 85% to 95% of the content that resides in our ecosystem today does not get translated because the business case for such content does not justify the cost.
A good example is a support knowledge base that happens to be there for customer use by, say, an electronics manufacturer. Generally the ROI required to translate all of that content into the 30 or 40 languages you do business in doesn’t quite exist. Unlike the marketing brochures for that same product, blogs and customer created content shelf life is short. So the many business drivers for funding traditional localization are in the top of the pyramid.
Next below that top tier on the pyramid is what we call just-in-time content. This is a large portion of today’s enterprise content, 85% to 95%, that does not get translated because of the ROI and the changing nature of the content.
This solution with IBM that we’re talking about is introducing an answer to the just-in-time content that needs to be translated from a source language into any target language. This is what we’re bringing to market now with our relationship with IBM.
So back to your question, the reason why we’re seeing new prospects is that generally the market had always ruled out the possibility of human translation as the only way to solve the problem and therefore given up in certain cases. Certain customers have accepted the fact that translation is expensive, it’s human with technology and there is nothing like real time translation that can solve this problem. However, with this announcement, they are now starting to scratch their heads and say; Hmm, maybe an application that I have such as Support FAQ’s, Knowledge Bases, Front End Applications, e-mail and IM chat within the enterprise can now be catered to and solved with this real time translation solution. We’re seeing a whole category of new customers coming to us with interesting questions that many thought earlier were not addressable with the old way of doing things through human translation. I hope that answers your question.
CSN - It absolutely does and it flows perfectly into my next question. In comparison, Google and Microsoft are also addressing this next tier content bracket with free services. What is the unique advantage or differentiator of your new joint product with IBM over those free services?
Satish - That’s a great question. Yes you are correct, there are existing consumer level translation capabilities available for free from Google and Microsoft, and this is by no means trying to compete or introduce another element in the market for that use case.
What we are clearly focused on is the enterprise as opposed to the consumer. The enterprise has a very different set of real-time translation needs. Quality expectation, while good enough is sufficient for them for a certain category of applications in this just-in-time category, they clearly would prefer to have the ability to get better quality by using their own language models.
A good example I can give you is an enterprise that is specializing in baby powder. They would prefer to have their real-time content customized and tailored with the terminology and language model that is specific to their enterprise. You can extrapolate that to consumer electronics, printer manufacturers, etc., you get the idea. Our ability to bring real time translation solutions to market in a SaaS hosted, multi-tenancy model is what differentiates us from the Google and Microsoft offerings. The same enterprise customer who I referred to earlier will today not be able to customize their language model by using the Google or Microsoft solutions. Those are one size fits all consumer level, free, MT solutions.
What we’re trying to bring to market is a solution for the enterprise customers to tune their respective tenancies with their own language assets as opposed to diluting it with someone else’s language model and therefore drive better security, quality, and customization which is key to the success of our solution. Very different, in concept it may appear to be the same but as you can tell the deployment, security, and customization models are entirely different for the enterprise, two different applications of technology.
We have found that by training the engines very specifically with the language models of a given customer and by further defining it based on product specific language models for that particular project the quality gets better, and better, and better. So if you’re an enterprise and you’ve got your support site for your consumer electronics offering you would be much better served by having that site real-time translated with its own tenancy, driven by your own language assets. The quality output is higher and you have control over your tenancy as opposed to trying a one size fits all MT solution.
CSN - Is there any cross sharing of those customer-specific language assets with IBM?
Satish - No, clearly no. In this relationship Lionbridge will be hosting all the technology and the customers assets just like we do today in our services offerings. We today manage a tremendous amount of assets, translation memories I should say, for our customers and we do that in a multi-tenancy mode in our translation work space platform. So at no point would those customer assets be accessible or turned over to IBM. IBM is our technology partner, not our go to market execution partner in this case. We are hosting all the assets for the customer securely.
CSN - In terms of performance, have you benchmarked performance quality advantages for your just-in-time tier solution?
Satish - Well, the quality expectation for the good enough, just-in-time application is very different from that of your high-end, just-in-case applications so the benchmarks tend to be quite different. One is business scale quality. If you’re writing a marketing brochure for a high end product for the next digital camera your expectations for quality are quite high on that piece of literature. Or, if you’re launching the next flagship platform product the software localization absolutely has to be top notch. For that you have a very different set of quality expectations on the benchmark. We have done some very exhaustive tests with the market and what’s out there. The one thing I can say without giving specific numbers away is that this real time quality, for real time translation, will be a notch below human, high end quality. This industry segment accepts that, but I do have to say that it is significantly better than the generic MT engines that are out there where you don’t have the ability to customize your language models for each customer’s tenancy.
I do want to emphasize that there are two different quality notions here; one is high end quality for just-in-case translation which is done at a high premium cost and then there is acceptable, utility level, multilingual communications where you know that you’ve got the general “gist of” the translation of that FAQ in Chinese or what have you. You’re not sweating it if you don’t have 100% quality because you just got the gist of what the problem was through real time translation.
CSN - Let’s talk about product deployment. I know that IBM has been utilizing this technology internally for about four to five years as it was being developed.
Satish - Yes, actually IBM has been investing in this technology in various incarnations of speech and text related technologies for almost 30 to 35 years so they’ve got a very good foundation through multiple generations of engines that they’ve built. With this technology they spent quite a bit of money in research and hardened this particular version of it over the last five to seven years. They actually brought this technology to the front given the recent innovation contests that they’ve had within IBM. They have also deployed it internally within IBM in several use cases over the last few years. So we’re pretty confident with the stability of the technology and the fact that we’ve done pretty extensive tests prior to arriving at this partnership ourselves.
CSN - How optimistic are you at this point there you’ll be able to fully deploy by the end of this year or early 2011?
Satish - Very confident, we’re confident with the stability of the technology and deployment models and the fact that we, IBM and Lionbridge, pushed the boundaries of this technology as far as we can. What we’re more interested in learning is the deployment scenarios as the customers see it. We want to make sure we understand that quite well by working closely with several large pioneer customers and then bring that to market. At this point, it’s really more about the deployment experience we learn listening to our customers. We want to give ourselves that time between now and the end of the year.
CSN - As we discussed earlier, IBM has been involved with this technology for a very long time, a pioneer in fact, but they seem to have been dormant for some time. What has occurred in terms of either market awareness or perhaps Lionbridge’s influence to reignite or refocus IBM’s attention on developing their MT product?
Satish - Yes that’s a pretty good question. It’s all about timing isn’t it? I guess you’re asking why now? What transpired to bring this to market now? It’s nothing specifically to do with IBM or Lionbridge. There is an element to that but I will come back to it. More importantly, with the ubiquity of the web, the mobility of content, our ability to create, access, and distribute content from anywhere, anytime, on the go, the appetite for folks like you and I as consumers to want to get anything on demand at any given time is now a lot more than what it was perhaps just a few years ago and it’s continuing to rise on an exponential basis. The patience level of sitting there waiting for something to be pre-translated just-in-case for certain categories of content is just not there today.
I could see our generation Y coming up, working from anywhere on their laptops or iPads basically saying “I want that piece of content available in Spanish now.” And they’re willing to accept good enough communication. So the demand, based on mobility and ubiquity, I think is clear and present today.
As a result of that what we find that folks who are putting content together, whether it’s for the enterprise, the content providers, or whoever, are now stuck with this problem of enabling real-time content as quickly as possible.
Whenever I engage with my kids through IM it surprises me that they can actually pass an English test. But guess what, they’re perfectly OK with good enough quality translation, in their own words, just to get the point across and we’re seeing that for a certain class of content that it is also OK and you don’t need publication grade quality.
I think the time has jelled with the notion of mobility and ubiquitous access to content anywhere and anytime. I’m sure you would agree that the time is now, even though the technology has existed in shape and form for many years.
CSN - With texting and IM so prevalent now, just good enough communication has practically become a new language.
Satish - Exactly, the amount of typos that come through is stunning however the messages get across, nobody goes back and corrects them and says “sorry I mistyped that.” Instead they know you got the idea.
That’s exactly what we’re talking about. So quite often Keith, what happens is that machine translation is sort of bucketed as an enabler to the high-end quality of translation. There are applications of machine translation in that category as well. What we’re seeing, and that’s why this announcement is very different, is that this is not just about using MT for the high-end quality of just-in-case translation, this is enabling a whole other ecosystem of real-time communications. It’s ubiquitous, mobile communication where good enough is truly good enough.
CSN - So, with this new solution, you’re not competing with the current load of translation business, you’re actually working to develop an untapped segment of the market.
Satish - Exactly, we’re giving testimony to a whole new deployment scenario of technology where the problem hasn’t been addressed before. I’ll give you some examples of numbers as well. Depending on whose numbers you believe, the industry market figure’s say the rough number of qualified, tested translators in the world is roughly 300,000 to 450,000. The average productivity of a translator is about 2000 words per day. So it’s a fairly limited pool of translators that are out there, right? It sounds like a lot but at the end of the day the amount of content that translators can translate is not a whole lot of capacity.
So what we’re trying to do is expand the viability of bringing technology solutions, real time translation in this case, to a much broader pool of content that today is not served by the high-end human translation because the ROI doesn’t make sense, resource availability is limited, or the content changes so quickly.
This is truly a new and different market.
Published - November 2010
ClientSide News Magazine - www.clientsidenews.com
Please see some ads as well as other content from TranslationDirectory.com: