MT: Future of eEverything, or Precursor to Unemployment?
By: Wayne Bourland (Dell) - Dell Computer Corporation
24 November 2014
If machine translation has already proven to be so great, why are many companies still so slow to add MT to their workflow? The paradigm of “cost x quality x velocity" is a key to the future of our industry. In this article, Wayne talks about the role MT plays in reaching the desired balance between these three components as well as the risks and the challenges of adoption.
We keep hearing that MT is the future of localizing eEverything, but if MT is the future of eEverything, why is eNobody catching on? Dell was at the forefront of using post-edited Machine Translation [PEMT] for eCommerce, and while that was years ago, we are still at the forefront, and that’s a big problem—for me, and for the industry. The cutting edge is only supposed to be cutting for a short period of time before it becomes the norm and something or someone else leapfrogs it. Unfortunately, our clinging to outdated quality expectations is holding us back.
Machine Translation [MT] quality isn’t as good as human translations. That’s an important statement to make right up front, because people need to accept it so they can move on! It will be a very long time before MT quality is as good as human quality, and if you wait until then, your users, clients, and consumers will have left you long before you’re ready to adopt MT. Gen Z doesn’t care about proper spelling, grammar, or really much of anything else, except getting the information they want, when they want it, the way they want to consume it.
Perhaps you have an unlimited revenue stream and you can afford the best quality translations for all of your content. If so, you can stop reading now. Unless of course, you need to deliver that content in a few hours. No human translation process can match the velocity of MT, not when a few servers can spit out millions of words of somewhat good translations in the time it takes a human translator to download the job. George S. Patton said, and I paraphrase, “a good plan now is better than a perfect plan next week.” I am pretty sure Patton didn’t know anything about translations, but he did understand the value of velocity and the concept of utility over aesthetics. Gen Z would be a fan of Patton, if they knew who he was.
Velocity and Utility….Velocitility!
So, if velocitility is the new quality paradigm, why don’t we just use raw MT for everything, replace our translation departments with APIs, and use some of our savings to help retrain translators for a different career field? Answer: because quality still matters. Gen Z isn’t spending a lot of money yet, and people still have an expectation of quality. That expectation is just different from what traditional translation departments have lived by for decades. We need velocitility at the right level of quality; a quality threshold set by the consumers of our content instead of linguistic standards and marketing departments.
Some content can leverage raw MT, and there is still a need for pure human translations, although that space shrinks as technology improvements and consumer desires merge at an apex significantly lower than the quality output of copywriting translations. The vast majority of content, the sweet spot really, can be addressed with post-edited MT. PEMT is the best of both worlds; you get velocitility as well as a human touch to meet consumers’ changing quality demands. It’s not going to be as good as what you’re used to, but that’s ok. It will cost less, be delivered faster, and will meet the needs of your consumers. And what’s more, it will allow you to deliver value to the enterprise where it counts, the bottom line.
The dam is going to break at some point, and we will witness a flood of MT adoption. Not in the traditional space of support content and product documents, but across the content ecosystem. The question is, will that happen before I lose my job? You see, there is a danger to being on the cutting edge; you set unrealistic expectations for the future. My boss recently asked me what we are going to do next, what is the next big innovation to refine the cost x quality x velocity = results equation? Right now, with so little adoption and fear of moving away from tried and true quality standards, MT, and the localization industry as a whole, has stagnated.
To reinvigorate innovation in our industry, we have to accept risk. When we first launched PEMT for product content on Dell.com, what seems like ages ago, we did not tell anyone what we were doing. We pushed the button and sat back, cringing, waiting for the complaints to flood in. Two months later, having received not a single escalation, we went full throttle, rolling out PEMT for all 28 of our languages within a year.
In truth, I didn’t really cringe. I was confident we were doing the right thing. We had listened to our customers. We knew accomplishing their purpose on the site was more important to them than perfect linguistics, and we knew we had a good PEMT model in place with vendors we trusted. Nothing worthwhile or innovative can happen without taking some risk. If it is a safe bet, you can bet it isn’t innovative. I am not talking about the kind of risk that brings down a multi-billion dollar commerce site or shuts off the Los Angeles power grid, but rather the kind of risk that allows you to cut costs by 25% and increase velocity of translated content by 50%. It starts with understanding your end customer, who they are, what they expect, and how those expectations are changing. We accomplished this with usability studies and surveys. For some, getting customer insights may be challenging, but it’s worth the effort.
In the end, the only risk that can hurt you is the one you don’t take. If you’re doing the right thing for the customer and the business, nobody is going to fault a broken link or a minor translation transgression. However, if we as industry thought leaders don’t step out and move the collective innovation threshold of the industry, we are all at risk of being bypassed.
Wayne Bourland is recognized in both the content management and localization industries as an agent for change, driving innovation and process efficiencies across global organizations. After a decade-long career in the US Army, he joined Dell, starting as a rep in the call center and quickly moving to managing call centers, launching call centers globally, and then into content management and localization. He is currently responsible for translation of Dell.com, support content, learning and development and marketing collateral for more than 80 organizations across Dell.