UK-based translation service provider SDL plc has announced it will acquire statistical machine translation (MT) vendor Language Weaver for US $42.5m. Why does SDL want to pay 3.5 times revenue ($12.2m in 2009) for a money-losing ($1.0m in 2009) technology company? According to one active blog-o-tweeter (who coincidentally represents an SDL competitor), it's a cynical ploy to impress the London stock market by demonstrating that SDL is becoming a software company rather than a services provider.
Well, ฉันคิดว่าไม่ได้. (That's Thai for "I think not," according to a free translation powered by Language Weaver.) SDL wants Language Weaver because they correctly realize that there is a huge future opportunity for machine translation. This combination makes sense for SDL because it solidifies their translation value chain (MT plus translation memory databases plus humans plus WCM plugins). And it will impact Content and Collaboration professionals because:
- The internet makes machine translation inevitable. The web produces an ever-growing perfect storm of massive amounts of content combined with millions of consumers who want to interact with it in their native language. Human translation simply doesn't have the capacity to do the job. Bottom line: The internet is now so pervasive that it demands machine translation.
- Raw output from machine translations is playing a larger role on web sites. Blogs, knowledge bases, chat, and some kinds of user-generated content are candidates for raw MT. Trip Advisor, for example, user Language Weaver translations of traveler's reviews.
- MT plus post-editing reduces the cost of human-quality translations. Machine output with human review and refinement can be significantly cheaper than human-only — allowing organizations to produce more localized content within a given budget and/or to localize content that would otherwise have been too expensive.
- MT drives global business and global business drives MT. "International operations" is no longer a prerequisite for being a "global business." Say you run a small hotel chain in Western Europe. How many of your guests come from Eastern Europe? Russia? China? How do you communicate with them? According to Common Sense Advisory, only 15% of consumers are very comfortable buying online in a non-native language.
Why haven't more enterprises already embraced machine translation? Vendors cite two leading responses:
- "It's not good enough." The question is, good enough for what? For decades after its invention during the Cold War, MT labored under the delusion that machines could produce fully automated human-quality translations. (A leading criteria for judging the quality of machine out put was in fact called FAHQT.) But business pressures in the internet environment help organizations see that the question is often not, "Is it flawless?" it's "Is it useful?" It would be silly, for example, for Trip Advisor to pay for humans to translate the reviews that populate their sites.
- "I can get it from Google for free." (Unsolicited hint to the masses on behalf on another giant company: You can get it from Microsoft for free, too.) The question is, what is "it." Google Translate and Microsoft's Bing Translator are smart and powerful statistical machine translation engines. (Google Translate, Language Weaver, and Asia Online all began in the same academic work.) So why would/should a company pay for an SMT solution? SDL says companies like the "security and confidentiality" of controling their own content. But the better reason is that, unlike Google and Bing, the commercial solutions are trainable. They can learn domain-specific vocabularies and improve over time, especially in conjunction with a translation memory system. (Which means I won't be too worried if ฉันคิดว่าไม่ได้ isn't a perfect translation of "I think not.")
Finally, don't forget that SDL is the mothership of WCM vendor Tridion, which it acquired in 2007. (Now called SDL Tridion.) In the podcast accompanying the announcement, SDL CEO Mark Lancaster stressed the value of further enhancing Tridion's already impressive multilingual content management capabilities with the addition of SMT. That's true, but it might be like adding another layer of wax to the Ferrari. As WCM turns into Multichannel Customer Engagement Management, SDL Tridion will be better served by polishing its positioning and messaging for marketers. (Wait: Can MCEM fly as the much needed acronym replacement for WCM? Answer: ฉันคิดว่าไม่ได้.)
What do you think? Are you using MT (with or without post-editing) now? Do you plan to? If not, why not?