The notion that translators are an endangered species has already been circulating for some time. Machine translation (MT) offered by companies such as Google or Bing is, after all, free of charge and available any time. Granted: MT is improving year on year. It is fairly adequate for private purposes, for example when you are abroad and need quick help with directions or a hotel booking. Even if you cannot rely fully on the offered translations, communication is simplified.
Professional translators do not use such machine translation, i.e. universal electronic translators such as those by Google or Bing, but only computer-assisted translation tools, which make it possible to manage databases and maintain particular text segments. In fact, computer-aided translation tools are essential when it comes to maintaining terminology consistency. You can use a translation memory (TM) for as many specialisms, clients or language combinations as you want. A TM designed for technical texts will recognise that a "crane" is not a bird in the given context, but a machine used to lift and lower materials.
There are quite a few reasons that MT will not replace human translators in the near future:
- No contextuality
No translation tool can know the intention, context and purpose of a particular content. Any wishes or preferences of the client will also not be taken into account.
- No originality
Nobody wants an ad that is full of the same terms as the competitor's. A "universal translator" can't take into account any corporate terminologies, slogans or other such contents.
- No legal certainty
It doesn’t bear thinking about what can happen when MT is used ad hoc during court proceedings, or when you resort to MT whilst translating an important contract. Who do you want to sue if a million-loss is incurred? And: what guarantee do you have that your confidential document will not be used and processed by the MT provider?
- No quality assurance
Normally, MT makes use of pre-existing human translations, selecting the best matches with the help of statistical methods. However, MT can't evaluate the quality of various translation proposals in a given context. How many bad translations do you find on the internet? Who will proofread the MT-rendered translation and compare it with the source text if you decide not to use the services of a human translator?
- No detection of substantial errors possible
As a translator, I often come across incomprehensible expressions or careless mistakes in source texts. Some printed documents also feature marginal notes or erasures. A translation tool is not able to take into account such special cases, let alone adequately evaluate their significance.