Monthly Archive: May 2014
A client recently sent me a translation request for a technical user's manual where we had already worked on another manual for the same manufacturer. The client put the translation memory and termbase from our first delivery through a machine translation solution to translate the second manual. He then asked us to "post-MT edit" the new job, with the expectation (or at least, hope) that we'd be able to offer a lower price for the same quality of work.
Here is my response to him after checking with my team.
"I discussed this with my team; I even presented the machine translation in as positive a light as I could (emphasizing how it leverages our TM and terminology and isn't just raw out of Google Translate) and simply asked what rate we'd need to charge to do it (without suggesting a steep discount).My colleague acknowledged that there are a few segments where the machine translation can help, but by the time he puts in the effort to understand the English and sort through the many more segments that don't help at all and/or to re-craft the sentences around a natural writing style, the result is that the machine translation is no help at all, and is actually a hindrance if using it for anything more than terminology mining (which we can do in memoQ anyway without bringing in MT).His feedback mirrors my opinion on a project I'm working on now where the client actually provided me with the translation of a previous version of the same document done by another human translator to use as reference. Even on segments with 100% matching, I don't think the other translation helps at all if I'm expected to deliver my best work. That's because I still have to understand the source and put together an English sentence that I am satisfied with. Even where the other translator's translation is fine, it's usually not the phrasing I would have used and/or doesn't match the style I used to translate other segments from scratch, not to mention various mistranslations that need correcting. The result is that the existing translation is of marginal, if any, help to me.Furthermore, I don't know how other translators see it, but for me, building onto an existing translation is unpleasant work compared with having a clean slate, so everything else being equal, I'd still rather just handle a regular translation job. I suspect that this because the act of composing and reworking my translation is part of the process I go through to understand the source text deeply and that having a pre-existing translated text interferes in this exercise, forcing me to work in a way that feels constraining. I also prefer an approach that lets me go through a document with a draft first and then come back for a second editing to correct, but I find that I'm not able to do this when I have to press my translation into an existing text, once again pushing me out of my comfort zone. It would be interesting to find out if I'm unique in this regard or if this is a common translation mindset.Working off a machine translation would be even more challenging than this. Therefore, if the client's objective is to receive work on par with what we usually deliver, then I would only be able to handle that at standard rates (applying volume and gold text discounts).At least for the English/Korean language pair, I'd say that the only way to make post-MT editing work is if the client is willing to accept a "good enough" delivery. In this case, there'd be no point in an extra proofreading step or in using subject-matter experts like we usually do. The translator would only need to focus on delivering something understandable and not have to worry about being held accountable for a publishable final version. This is still more than MT often delivers, but is not what our clients usually pay us to provide.I'd be willing to check with another resource for this kind of post-MT edit workflow and target a price at about half of our standard rate. Would you like me to try that and get back to you on it tomorrow?