Anyone working in the translation industry has probably heard friends say "In a few years you won't need translators anymore, the machines will soon be able to do it better". Not only does this sound sobering to the professional, but it often unsettles the customer who needs and orders translations. Here are a few thoughts on this topic.
Yes, automation is advancing all the time, and today machines are able to replace humans in many tasks. Horses have experienced this before: In 1888, Bertha Benz - Carl Benz’s wife - together with her sons undertook a secret journey from Mannheim to Pforzheim to test the first "horseless" carriage. In Wiesloch the tank was already empty. Now what? Of course, there were no gas stations yet. The fuel for the onward journey - in the form of Ligroin - was supplied by the city pharmacy, which thus achieved fame as the "first filling station in the world".
Back to our work: machine translation will certainly take over some of the tasks of translation, but it will also create new tasks.
For laypersons, it may initially be important to distinguish between computer-assisted and machine translation. Computer-assisted translation is carried out by means of a computer program (CAT tool - computer-assisted translation tool), which in principle makes it possible to make the work of the translator and the reviewer easier and more efficient. The main feature of a CAT tool is that the user has access to a database where previous translations are stored. This makes it possible, for example, to look up a technical term that was agreed with the same customer when translating a previous user manual so that it can still be used with the same meaning. We call this aspect "consistency". For example, if a command interface is sometimes referred to as a "button" and sometimes as a "key" in a user manual, the manual lacks consistency – and the reader does not realize that it is the same command interface every time.
Machine translation goes a step further: the algorithms of the computer program are able to simulate the processes of a human brain learning a new language. This is done using a huge dictionary that consists of all the translations available online. And just as people use many different learning strategies, machine translation can be based on different methods. Examples are rule-based methods for machine translation as well as neural, statistical and example-based machine translation. Depending on the type of text, a particular method will produce more or less successful results. In some areas, machine translation can already serve as a useful pre-translation, which is then processed "humanly".
To what extent do these possibilities change the work of translators and reviewers? If the machine becomes more intelligent, people will also be expected to have fewer muscles and more brains. One advantage of a machine pre-translated text is that the translator does not have to type all the words. But the energy saved is needed to check and revise the text carefully, understand the context in depth, discover possible pitfalls regarding technical terms, as well as to compare reference material. As a result, the translator's performance no longer depends on typing speed, but rather on concentration, excellent language skills, a good eye for key terms and technical understanding. This activity is called "post-editing" and should be learned specifically, for example by attending appropriate training courses.
Ultimately, it is still up to the translator to decide whether and when such machine support may be appropriate. For some texts, the program works like an untalented translator: instead of revising the result, it is just better to translate everything from scratch.
One of the reasons we love our work is because it is diverse. That is why we are also looking forward to this change with confidence. Curious and open-minded, we continually educate ourselves about the prospects of computer-assisted and machine translation, so that they do not become and remain competitors but helpful tools.
Sections