Hello everyone! My name is Chris, and I work as a J-E LS for Gengo. I have been a freelance translator for something close to 15 years at this point, and have worked in a fairly wide variety of sectors by now, getting to see the industry evolve over the years.
This month, I have been lucky enough to get a chance to discuss an important translation point with everyone, and wanted to go over machine translation (MT) and post-editing (PE) for it. More than anything else, this issue has changed drastically over my time as a translator.
To begin with, I am sure everyone know Gengo does not encourage MT usage, and this is a view I agree with. That said, especially for the past couple of years and likely more so going forward, we can all expect to encounter cases where clients request MTPE for one reason or another. On that front, I thought I would go over a fairly compact example to lay out some pointers on an effective approach to MTPE (assuming a sort of “standard” project).
To begin with, speed is a central point in MTPE as a work type. Accordingly, the first step to post-editing a line is to rapidly read the target language only. I have included a pair of lines from a product listing to illustrate.
落ち着いた色で、どんな食器にも合わせやすい天然木製スプーンセット、飽きのこないデザインで、長く愛着を持ってお使いいただけます。
Natural wooden spoon set in calm colors, easy to match with any tableware, with a design that will never get old and will be loved for a long time.
使用範囲：家庭用、レストラン用、ハイキング用、弁当箱用、通勤、通学、※デザート、ケーキ、パスタなど使います、家でお気軽にカフェ雰囲気を味わえますよ
Scope of use: home use, restaurant use, hiking, lunch box use, commuting to work, school, * dessert, cake, pasta, etc., you can feel free to enjoy the cafe atmosphere at home!
If a translation is incomprehensible, the next step would be to delete it and start an original translation from scratch. However, cases like the above are more involved, and the next step involves noting any issues with the flow or grammar of the target text (shown with bold formatting here) and starting a close review against the source.
When we do this, it is a good idea to check more closely anywhere that seemed unnatural in the target language, as well as any idiomatic phrases, numbers, or things like long or complex sentences. And as it turns out, these are the causes for several of the points I highlighted with bold formatting from a read-through of the target English.
To highlight specific points, we can see where the MT slips up on the mildly tricky term お気軽に at the end of the second line. The term can sometimes mean “feel free,” but here the MT is obviously using the match in a way that doesn’t fit in context. Typically, MT struggles with situational translations like this and the term 愛着 in the first sentence (around the second bold highlight).
Other prominent issues include repetitions (be sure to watch for the MT repeating things it doesn’t need to, like “use” in the second line, or repeating words that do not repeat in the source, separately), word structure disagreements (at the start of the first line), target language inconsistencies (in the list in the second line), and issues with use of articles. In another format, I would show this with an edit-by-edit approach, but here I will have to rely on a before and after, with the edited English below.
Natural wooden spoon, in subdued colors and easy to match with any tableware, with a design that will never get old, making it sure to be a favorite for a long time.
Scope of use: home, restaurant, hiking, lunch boxes, commuting, school, *Dessert, cake, pasta, etc.; a convenient way to enjoy cafe atmosphere at home!
And that covers everything I can fit in here! Of course, even with all of these points laid out, these edits are still partially subjective – but that is something that will always be a part of translation to some small degree, even when working with MT. Of course, I also know this was fairly limited due to space, so please also feel free to chime in or ask about anything I haven’t been able to get to!
0 comments