Is translation memory obsolete?
CSN recently learned of a new product from Lingotek. We wondered if this technology would make translation memory, as we know it, obsolete. To learn more about the product and the technology behind it, we interviewed the CEO of Lingotek, Tim Hunt.
CSN – When we first looked at this product, we couldn’t help but draw a comparison to translation memory (TM) tools. Can you briefly define the product and explain why it’s not just another TM tool?
Tim – Translation is about knowledge. The best language professionals are the ones with the most knowledge. TM tools provide language professionals with a photographic memory of previously translated sentences. However, this is not enough; there is more knowledge in existing translations than repeat sentences.
Sure, Lingotek can provide the repeat sentences, so you might think it is another TM tool, but it also opens up all content in a document for reuse. And I do mean all.
Can you imagine using a TM tool to translate a document when you know up front that you likely won’t get even one exact-match sentence? With Lingotek, you won’t hesitate. As the content grows, it is hard to find a paragraph where you don’t get hits. Lingotek’s new language search engine increases content reuse more than 80 percent over traditional TM.
CSN – Lionbridge said in july that centralizing their individual tms and providing access to them over the internet was increasing content reuse 5 to 10 percent. do you really think you can increase content reuse to 80 percent?
Tim – Yes. And our content is centralized over the Internet also, so some of our reuse gains come from better access, but most gains come from the way we index and search content. In 1993, Pierre Isabelle of Canada stated that “existing translations contain more solutions to more translation problems than any other available resource.” The problem is how to get the “solutions” out of the “existing translations.” TM did it with sentences, but sentences don’t repeat enough to use TM tools on all translations.
Lingotek has taken the technology of the Internet search engine and added to it meaning-based searches. The result is a tool that provides translations that a TM can’t provide. And it doesn’t require the user to do anything with the data to get the content out. They just translate. We eliminate translation memory management.
CSN – Without getting too technical, help us understand how you use mainstream search engine technology to create a language search engine?
Tim – Frankly, TM tools are designed wrong. Just think if we had to search the Internet with the TM approach. You would find hits only if the sentences were a close enough fuzzy match or an exact match. How many people would use an Internet search engine if it regularly told them “no match.”
But you can’t just use the Internet search engine “as is.” Internet search engines do concordant-type searches, which provide millions of hits per search. With such results, a translator would have too much content to sort through. And an Internet search engine doesn’t know if a word means anything different in a different context. Translators need to make sure that the “hits” they get use the same words with the same meanings in the same contexts. Only then are the translations relevant.
That is what is at the core of our product’s search. We added meaning-based search capability to the fast, large searching capabilities of the Internet search engines. I can’t tell you how that works because it is our secret sauce, but the language search engine distinguishes between the different meanings of a word in different contexts and provides the most relevant hits. This meaning-based search works for all the world’s languages. This is what will make TM tools, as we know them, obsolete.
CSN – This is really a bold statement. Large companies like SDL TRADOS and MultiCorpora have invested millions of dollars in TM technology. They are delivering thousands of units to the market, successfully fulfilling the market’s needs. Are you saying they all need to rethink what they are doing, including the user base of these products?
Tim – Yes. We can’t expect to get tremendous performance gains if we keep doing what we have done in the past. TM provided a value, one that we haven’t lost with the language search engine, but it is time that a technology was developed that lets a translator do in an hour what they previously did in a day.
But with a new technology, not everything will work like it did before. For example, when people went from VHS to DVD, they had to accept some changes. New technologies always bring change. And users ultimately have to decide if the performance enhancements are worth the effort to change.
For example, instead of installing software on their computers, users will have to get accustomed to translating online, using a Web browser tool that is software-as-a-service. Users will have to grant people access to their translated indexes in order to share content, instead of e-mailing or mailing all the data to translators or suppliers. Users will have to change some business practices and pricing when translators start getting translations done 200 to 800 percent faster. Sure, the translators will make more money per hour, initially, but eventually, competition will drive down word rates. With lower rates, clients will be able to afford to have more translations done. Those who use Lingotek will have an advantage in pricing and quality that their competition won’t have.
CSN – The big question, then, is whether or not Language Search Engines will replace TM systems. How do you think that will happen?
Tim – I think it can and will happen. There might be users in the short run who use both the TM tools and the Lingotek language search engine together. We see a similar model with video players that have both VHS and DVD. In the end, however, I think the performance gains of the language search engine will win out.
For example, if a translator was making $0.05 per word and translating at 2,000 to 3,000 words per day, they are working very long hours just to make $100 to $150 a day. And that same translator might not have a translation job every day. They likely have to work long hours on some days to make up for the time they don’t have work.
When that same translator increases his or her speed to 4,000 or 16,000 words per day, their pay increases to $200 to $800 per day. Now you might say that such an estimate is too high of an increase, and it isn’t possible. But if you think about it, software for accountants, graphic designers, and many other professions have allowed them to get 800 percent performance gains. Why not translators?
CSN – Are you sure? 800 percent performance gain seems too much?
Well, just so there is no confusion about the performance gains of Lingotek, we have built into the language search engine a meter that tells the translator how fast they are translating in words per day. I have seen users consistently getting rates of 8,000 to 12,000 words per day. I have seen it jump as high as 28,000, when the translator gets a lot of reusable content.
Even if we only double their speed, translators only have to work about two hours of each month to pay for the Lingotek service. The rest of their performance gains are money in their pockets. Likewise, translators wouldn’t have to translate very long before deciding they could compete for more jobs by lowering their per-word rates. If they find that they are translating at 8,000 words per day, they can still double their income, even after reducing their per-word rates. And they could do this while working more regular hours.
Likewise, language service providers (LSPs) would get lower rates from the freelance translators and could also become more competitive, taking clients from other companies that can’t offer the same pricing.
CSN – The market is flooded with TM users, and anyone in this industry can tell you that conversion to a new technology is a hard sale. What is it about your product that makes you feel you can compete with the biggest companies in this arena and unseat a trusted technology like TM?
Tim – Your right, it won’t be easy, and it will likely take time. But we are not trying to compete with the TM tools. Instead, we are out to change the industry. How that change will come about, exactly, we are unsure. Perhaps we are making too bold a move, but even some of our first clients have recognized that this is a disruptive technology that can change the industry.
Initially, some users will use Lingotek on some jobs and their old TM tool on other jobs. Some will be early adopters and do all their work with Lingotek. Others will create a login with our one-month free trial, and then never put any content into Lingotek to try it out.
We might even see some of the TM tool providers build their own language search engines and integrate them into their own tools. Some providers might approach us to enhance their offerings to their customers through integration with us. Since we are paid for the service, not the software, we would entertain offers to integrate with them. I don’t exactly know where the technology will go. But we built it. It is available now. And I’m excited to see what will happen next.
For more information about Lingotek, go to www.lingotek.com.
ClientSide News Magazine - www.clientsidenews.com
Please see some ads as well as other content from TranslationDirectory.com: