Selling Data

In 2004, I started a company with Donna Parrish, one of the co-owners and co-organizers of the LocWorld conferences (, called TM Marketplace. The idea behind the company was to facilitate the leasing of translation memories (TMs) between parties. These TMs could either be used for traditional TM lookup services or as training material for the then-blossoming statistical machine translation (MT) engines. The company didn’t survive very long, but I mention this here because I want to tell you a story connected with it.

One of the underlying ideas of allowing partners and even competitors to use TMs is that it really is in everyone’s interest. Localizing a first-of-its-kind product in a given language market involves a steep investment, especially with terminology that has to be researched and coined. The leasing of those materials to others might allow the leasing party to recoup some of that investment, but, more importantly, it might also give them a certain amount of power and prestige because their terminology would likely be used without highly company-specific or trademarked terms. Plus, because product owners believe their products compare favorably with the competition, it only makes sense to have everyone use similar terminology to create a more even playing field.

At a LocWorld conference in 2005, the heads of the translation divisions of SAP and Oracle were present. SAP had just announced they were about to release a Vietnamese version of their flagship product. While SAP and Oracle are very different companies, their core products are rather similar. (In fact, if you read through their Wikipedia pages, the description of their range of products is almost word-for-word the same).

Of course, I saw this as a great opportunity for TM Marketplace. It seemed a no-brainer to me that SAP would be more than willing to use our services so their data could be leased to Oracle and other partners and competitors.

I was essentially laughed out of the room by the SAP representative—who asked why in the world they would ever do that. The Oracle representative, while acknowledging that they were indeed gearing up for their own Vietnamese localization, also just shook his head at my naivety. That simply wasn’t how things were done!

Needless to say, I felt rather stupid, and, as you can see, I’ve certainly never forgotten the episode. But I’ve always felt that maybe I wasn’t the one who missed out on something there. (My reason for retelling this story will become apparent in a moment.)

I recently had an opportunity to speak to Thomas Wienold, head of the Language Experience Lab at SAP, about a tool that’s been around since 2017 but has seen some interesting development in just the past few weeks. The tool is called SAP Translation Hub (, a cloud-based solution that allows for the translation of documents and user interface files into 39 languages. It accomplishes this by sending files through a two-tiered process, where the tool searches within SAP’s “multilingual text repository” (MLTR) for a proposed translation and displays it along with a quality index so that the end user can decide if it’s a good translation and if they want to use it. The remaining strings are filled with MT suggestions that originate from SAP’s own neural MT engine, which has also been trained with its MLTR. The cost for all this is approximately the same as you would pay for other generic application programming interface-based MT engines (2.42 €, or USD 3.34 per 100,000 characters), but the expected quality is naturally significantly better. In other words, users can expect correct and consistent terminology.

Both SAP’s MT engine and its MLTR are updated constantly and reflect the assets that SAP uses for its own internal purposes. Furthermore, the MLTR makes heavy use of metadata, which allows for industry-specific employment of that data. This is something that can even be improved by the possibility of adding a company’s (here SAP’s) own bilingual data to train the MT engine even more specifically for their purposes.

Once processed by the system, users can then edit the strings in a simple and proprietary interface.

What’s new and makes this even more attractive to anyone working with SAP’s data is that it’s now also available via the XTM Cloud interface and licensed via the SAP Store (the company’s online marketplace where customers can discover, try, buy, and renew solutions from SAP and its trusted partners). The XTM Cloud interface provides access to SAP’s specifically trained MT engine, just like any other application programming interface-based MT engine, whereas the application offered in the SAP Store contains the MLTR lookup feature combined with SAP’s MT engine via the two-step process mentioned earlier supporting user interface translations. You can expect other translation environment tools to also offer this as a possibility in the future.

I asked Wienold whether he sees SAP Translation Hub as a “profit center” now or at some point in the future. He responded that he sees the tool more as an “enabler” for its users. The tool is open to anyone, including competitors, to which he has no objection. Nor should he, in my opinion. In fact, I think this is a fabulous solution that other companies could emulate, especially those like SAP with a lot of customers that develop their own extensions and products to fit into their respective ecosystems.

Plus, I just really love innovative ways to look at translation. And this is a prime example.

Jost Zetzsche is a translation industry and translation technology consultant. He is the author of Characters with Character: 50 Ways to Rekindle Your Love Affair with Language.

This column has two goals: to inform the community about technological advances and encourage the use and appreciation of technology among translation professionals.

The ATA Chronicle © 2023 All rights reserved.