Using Machine Translation for Translation and Audiovisual Translation, a Casestudy

For my master’s thesis, I carry out research on using machine translation for both the
translation of text and audiovisual translation. The aim of this research is to explore, describe, and
evaluate this translation tool to find an answer to the main research question: is machine translation
applicable to autonomously translate texts and audiovisual segments? By combining three different
methods – a literary review, a practical experiment, and an observation during my internship – the
aim is to implement methodological triangulation to be able to come to a conclusion that is wellfounded.
To examine machine translation for text and audiovisual content a division is made between
these two translation branches to properly carry out in-depth research per branch which is then
brought together to compare the results and draw a conclusion.
By comparing how machine translation is used for the two translation branches mentioned
above, it quickly becomes clear that the tool works in very different ways, depending on what kind of
content needs translating. This is because for the audiovisual translation an extra layer of complexity
is added as spoken language needs to be converted into written language before translation whereas
for the translation of written language this is not applicable. As it is a more complicated matter,
scientists have only started investigating it about thirty years ago even though the first research on
using machine translation for text stems from the 1930s. In addition to this, not many scientists have
ventured to do research on automated audiovisual translation which is why this practice is still
underdeveloped and very few sources can be found that show practical findings. This lack of research
and development makes using machine translation for audiovisual translation a very sophisticated
endeavour.
Not only the way machine translation works for the different translation branches differs, but
also the mistakes that can be found in its output. When evaluating the machine translation of text the
following main issues were found: consistency, meaning, and fluency. For the audiovisual translation,
the tool had more problems in regard to interpreting the spoken language correctly and completely,
and producing well-written and grammatically correct subtitles. This was mainly due to the nature of
spoken language, which is more colloquial and often contains linguistic structures that are not
acceptable in written language and therefore need to be rephrased in the subtitles. In general,
machine translation seemed to have issues in both translation branches with producing well-written
language that could have been written by a native speaker. All these mistakes machine translation
tools tend to commit show that machine translation can to this day not be used autonomously and
the interference of a professional and trained translator is necessary to guarantee a translation that
is qualitative.