Everyday, our industry becomes more and more competitive. Whether it be linguists competing for smaller projects, or agencies with large teams competing for larger ones, clients tend to look for the most economical, fastest, most reliable and highest quality outcome possible for their projects. But, as we know, that is not always the case.
In order to keep a client satisfied, language service providers are there to provide the best quality work to satisfy their needs. Translation agencies and companies tend to rely on 3 and sometimes 4 steps in order to achieve this. Nowadays, it may not be enough to provide only a 3-step translation service (we are talking about the standard Translation + Editing + Proofreading workflow). A fourth step is likely to be needed (depending on the kind of project), and that is a Quality Assessment step. For us, a QA exceeds a Project Manager’s capacity under certain conditions like when the word count surpasses around 10k words, when the PM is not proficient in the target language, whenever we deal with a lot of repetitions, or when our source file is a converted file (from scanned documents into editable Word files), just to name a few.
Let me share with you my personal experience performing this QA step. When dealing with large projects, often more than one linguist is involved, at least in the first stages of the project. The proofreader has to be aware of this situation so that he/she can focus on repetitions and segments that may have just a couple of words modified. If the editors involved can’t work together, or if there’s not a guideline in terms of specific terminology or phrasing to be used, then the proofreader will have a bit of a harder job altogether. Things can go sideways at this point, even though we are at the last stage of the project. I can say that sometimes (more than it should), we find that a few things are missing even after a 3-step workflow. We are humans, we can get tired after working for hours on the same project. Our eyes may be weary, and a few things may skip our focus. It is here where this last QA step becomes essential. We can rely on specific software to help us out through this stage. One of the best programs out there is Xbench, but not the only one as you can always count on built-in tools or other similar applications like Verifika.
But as I said, software is not all, and it is not always a breeze to perform a QA on certain projects. Two examples of factors that could give you a hard time are: a) projects where post-editing is used, and b) projects where our source is a converted file. In the case of MT, a good eye is needed to make sure there are no inconsistencies in our target language (this in terms of specific terminology), since the machine tends to provide different translations for the same word from segment to segment. This is especially true for the nowadays “in vogue” Neural MT.
For example, let’s take the Spanish word “laguna” (lagoon). In a project where the context/topic is the body of water, a machine would translate this word as “lagoon” in one segment, and then drastically change that translation into “gap” (laguna, in Spanish, also has that meaning when talking about a lapsus/mind gap). So, we can’t blame MT for not being consistent, since it has so far no way of determining whether or not the context requires the use of a synonym or a change in words. That’s where we still have room to claim for “added value” brought on by human beings!
Our other example is a converted source file. If we don’t do a good pre-editing job, right at the beginning, we might be doomed to encounter several inconsistencies in our project. From numbers converted into letters and vice versa, to misspelled words or “illegibles” that would make us lose context.
Providing a 4-step translation service can make all the difference. Just make sure every step is done well, right from the get go, so your QAer will only have to do the bare minimum. As the saying goes, an extra set of eyes means no surprise, so try to do as much as you can to keep your client satisfied.
