26th May 2021

Some background information

I'm Senior Vendor Specialist at Electronic Arts for the last four years now. Within the Vendor Management and Asset Loc Team – all based in Cologne – we handle all the translations and post-editing tasks of Electronic Arts. Vendor Management specifically takes care of recruitment, of managing our resource pool. We have about 130 vendors – single- and multi-language agencies and freelancers. So, yes, we take care of the trainings. We take care of the finance part and also of quality assurance, which is why we're talking today. We're defining our quality standards for post-editing, but also for translation.

You are so much involved in machine translation post-editing and you have a machine translation post-editing workflow in the company. Are there any other steps in the machine translation workflow, apart from the post-editing task itself, where you involve translators or post-editors?

We have a large pool of translation post-editing vendors, but we also have a large in-house testing team and linguistic experts. So while we do involve our vendors for the corpora selection, for example, when we try to create new engines, we rely mainly on our in-house staff when it comes to evaluating our engines. So, yes and no: our vendors do a part, and the other half is taken care of by our in-house staff.

Do you mean the steps before or after the post-editing itself?

Yes, we do have a post-editing workflow and the post-editing is being taken care of by our external resources, the corpora selection is also done by the vendors and the evaluation of the engines is done by our in-house staff.

Regarding your main job part, which is outsourcing and selection of vendors, etc., do you use translators as post-editors, or do you recruit post-editors based on the specific skills and competences? And if you use translators as post-editors, do you provide training or guidelines? And if you recruit post-editors based on the specific skills and competences, what are the skills and competences required in your opinion?

When we set out and started with machine translation post-editing, our aim was to keep as many of our translators as we could, and together with them, reshape them if possible into post-editors, because our translators and our reviewers are the ones who have all the project and gaming knowledge we need. As we post-edit in-game text as well as marketing content like metadata, store content or newsletters, the resources really have to have the knowledge. So for us it was clear we need to keep our translators. If we recruited everyone new we probably wouldn't have had a chance. We do provide them with guidelines, written guidelines for now. So when we started out, we created a lot of guides on what we expect and how they should handle the post-editing task, because most of them didn't have any expertise yet or any experience. And we are right now looking also into going a step further and also creating some kind of training online or through videos or whatever, but we want to take it further. And second part of the question, are we recruiting post-editors? Yes, whenever we can, but it is tough to find someone to have the combination of post-editing experience and the gaming knowledge. We're trying, but it is hard because the gaming knowledge and the knowledge on our products is what comes first for us.

For a gaming company, is it better to sacrifice post-editing in order to have the best gaming experience?

In a way, yes.

What are the quality expectations when post-editing and how do you assess post-editing quality?

We have different quality measures in place. We have like a big in-house pool, in-house resources, in-house testers and linguistic reviewers and we get regular feedback from them. Everything or almost everything we post-edit runs through the hands of a tester. We got weekly feedback from their side so that we know, OK, where are we? How are things going now? We have quality evaluations. That means we take random samples of our post-edited content, and we have some specific measures; well, metrics with which we then evaluate the quality of that, and we do project recaps. So whenever a game signed off (= reached the end of translation phase), we will also review the complete game, no matter whether it was post-edited or translated. And when it comes to expectations, our expectation there is that while you may always have this subjective feeling that something was post-edited and it may have had a MT source, we don't want our players and we don't want our readers of any marketing content to be able to tell. So basically, we apply the same quality metrics for translation as to post-editing.

As you said, you already provide training for post-editors in the form of guidelines and maybe you will have an online training that you will work more on based on this aspect. I suppose that what led your company to design a training for post-editors was the lack of training itself, the fact that there was not training from the university or that you would like to recruit gamers that did not have a post-editing experience and they were not, let's say, familiarized with the translation technology or is there any other reason?

As we are trying to keep our pool of resources and we just need to train them better and help them better, this is why we're trying to do this training. I know that some of our resources, especially the freelance translators, are already doing some trainings on their own from universities or from other sources. But we want to give them our view on things also so that they know how we do it. And especially this is how you could do it in our translation environment so that they know, OK, yes, I can do it like this. So they got some practical guidelines because right now everything is very theoretical, and we would just want to give them a hand to also do this better in practice.

Based on your experience, is this training efficient?

The guidelines so far, based on our scores, I would say yes, because right now all scores are quite good, so everything's fine, but I think we can do more. And this is what we are aiming at, to provide more and do more.

Is it difficult for the linguists to assimilate at some point the guidelines? They read, they study, but when in practice this becomes more difficult. Do you encounter this issue?