Half the money I spend on translation is wasted…

Half the money I spend on translation is wasted…

The trouble is, I don’t know which half.

The man in the picture is John Wanamaker, a nineteenth century American businessman, who is known for many famous quotes. Except for this one… his worry was that half of his advertising budget for his department stores was wasted. He couldn’t figure out which half. Despite the revolution of digital advertising and claims of innovative razor-sharp marketing strategies, today’s marketers still often refer back to John Wanamaker’s dilemma of more than hundred years ago. They know that a considerable share of their collective $650 billion spent on advertising every year (and growing fast) may still go down the drain.

Now we know

A CEO of one of the superagencies recently mentioned John Wanamaker’s famous quote to me and added that he felt that the same applied to the billions of dollars spent on post-editing machine-translated output in today’s translation industry. We know that half of these automatic translations are certainly of good enough quality and do not require a human eye or touch and edit. But can we tell which half? How do we know?

The good news is: now we know. With a new AI technology referred to as Quality Estimation or Quality Prediction, massive volumes of machine-translated content can be scored automatically on a quality scale. After a short learning curve and developing confidence in the models, the early adopters of this new technology in the localization industry quickly save 50% of cost and time by locking all segments with quality scores over a certain threshold from post-editing. And they aim higher: with MT getting better and the precision of the Quality Estimation models improving even further, 70% savings or higher may be achieved.

Deep Innovation

What counts more than the cost savings perhaps is the time we save to generate content in many more languages. ChatGPT has set the expectation at a level of real-time and massive multilinguality. Leading or following that norm requires deep innovation in the translation industry, causing perhaps even an identity crisis. Straker Translations has recently been rebranded as Straker Verify. The translation industry with its linguistic resources around the globe, the skills, and the language data is best placed to validate the output of AI and LLMs.?

Quality Estimation

Quality Estimation is an indispensable AI technology in every translation and globalization environment. It helps to save time and costs. What’s more, it changes the name of the game.

Learn more about Quality Estimation in our QE Guide. Or if you want to find out how it works on your content, check out our new Demo Interface.


Karen Hodgson

CEO Translationz

5 个月

Excellent information and will most definitely hit you up on your QE Guide.

回复
Eric Christiansen

IT Executive | Transformative, driven, and results focused with proven success leading technology strategy and operations for top-tier companies in finance, insurance, healthcare, and defense sectors.

5 个月

20 years ago we were able to remove non-value add processes through the use of the SAE J2450 metric. When GM found that warranty work based on technician understanding the documentation reached almost 0 with a J2450 score of 4 or lower, we were able to eliminate the post-editing process for our translation groups who were able to produce the quality needed first time. Same with our early MT implementations and we were able to score the output and determine what was fit for purpose vs what information could be efficiently revised vs what output was utter trash. It will take some effort for a system to be in place which also requires the customer to determine with some precision what their quality standards are.

Renato Beninatto

I help companies grow internationally by providing data, consulting and insights.

5 个月

Great analogy!

Godwin Josh

Co-Founder of Altrosyn and DIrector at CDTECH | Inventor | Manufacturer

5 个月

This shift in perception challenges the traditional value placed on post-editing as a necessary step. On a deeper level, this means questioning the very nature of "finished" content and the assumptions we make about its quality. So, are you seeing a rise in "pre-edited" content that bypasses traditional workflows altogether, and if so, how is this impacting the role of editors?

回复

要查看或添加评论,请登录

Jaap Van Der Meer的更多文章

  • What is the secret behind MT Quality Estimation?

    What is the secret behind MT Quality Estimation?

    At the start of this year I predicted that 2025 will be the year of MT Quality Estimation. We all know that 50% (or…

    9 条评论
  • 2025 will be the year of MT Quality Estimation

    2025 will be the year of MT Quality Estimation

    Market watchers predict the great normalization of AI for 2025. Time to reap the benefits: not just talk the talk, but…

    10 条评论
  • Looking Back on 20 Years of TAUS Where did we go wrong?

    Looking Back on 20 Years of TAUS Where did we go wrong?

    This month TAUS is celebrating its 20th anniversary. Before we launched the Quality Estimation and Human Language…

    15 条评论
  • Taos revisited

    Taos revisited

    I am back in Taos, fifteen and a half years after our first visit. What a beautiful and mystical place.

    1 条评论

社区洞察

其他会员也浏览了