Every day, our industry
is becoming more and more competitive.Whether it is linguists who compete for
smaller projects, or agencies with large teams competing for larger equipment,
customers tend to look for the result cheaper, faster, more reliable and of the
highest possible quality for your projects. But, as we know, that is
not always the case.
In order to keep a
customer satisfied, language services providers have to provide a job with the
best possible quality for your needs. Translation agencies and companies
often rely on three and sometimes four steps to achieve this. Today, may
not be sufficient to provide only a three-step translation service (we are
talking about the flow standard work of translation + editing + correction or
Proofreading). It is likely that you need a fourth step (depending on the
type of project), and that is a step of evaluation of quality, or QA. For
us, a QA exceeds the capacity of a Project Manager under certain
conditions, as when the word count exceeds 10,000 words, when the PM is not
competent in the target language, when we handle many repetitions or when our source
file is a file converted (from documents to editable Word files scanned), just
to name a few.\
When it comes to big
projects, there are often more than one linguist involved, at least in the
early stages of the project. The spell must be aware of this situation so
that you can focus on the repetitions and segments that may have only a couple
of words changed. If the involved editors cannot work together, or if
there are no guidelines on specific terminology, then the spell will have a job
a little more difficult. I can say that we find sometimes that there are
some things even after a 3-step workflow. We are human, we tired after
working for hours in the same project. Our eyes may be tired, and some
things can skip our approach. It is here where this last step in QA
becomes essential. We can rely on a specific software to help us with this
stage. One of the best programs is Bench, but not the only one, since you
can always count on tools built to each programmer or other similar applications
as Verifiable.
But as I said, software
is not everything, and is not always easy to carry out a quality control in
certain projects. Two examples of factors that might hamper it are: a)
projects that are utilize post-editing or machine translation (MT)
Edition, and b) projects that our source is a converted file. In the case
of the MT, you need a good eye to make sure there are no inconsistencies
in our language of destination (this in terms of specific terminology), since
the machine tends to provide different translations for the same word of segment
to segment. This is particularly true for the MT based on neural
networks.
For example, take the
word "lacuna". In a project where the context/subject is the
natural water tank, a machine would translate this word as "lagoon"
in a segment, and would then change drastically that translation to
"gap" (this time with the meaning of lapses).Therefore, we cannot
blame to machine translation for not being consistent, since up to now there
has been no way to determine if the context requires the use of a synonym or a
change in words or not. That's where I still have space to claim the
"added value" caused by humans!
Our other example is a
converted source file. If we do not do a good job of per-Edition, at the
beginning we would be condemned to find several inconsistencies in our project.From
numbers converted in letters and vice versa, until "illegible" or
misspelled words that would cause us to lose the context.
Provide a 4-step
translation service can make a difference. Only transfigure that every step
is done well from the outset, to make your concealer only to do as little as
possible. It is simply to do everything possible to keep your customer
satisfied.
