Texts are usually written to be read by people. However, as AI is being increasingly used to both create and translate texts, it is also very important that texts can be read by machines. If a text is to be machine-translated, there are a number of aspects that can be taken into account when creating a text to have a positive effect on the machine translation (MT) output. We will show you some potential stumbling blocks when using MT systems and potential solutions to significantly reduce sources of error and the amount of post-editing required for machine translation.

"Garbage in, garbage out?" was the title of the presentation that our Head of Quality Management, Eva-Maria Tillmann, gave last November at tekom (tcworld). Now the findings she presented are available to read and learn from. In the following article, she explains how working with computer-aided translation (CAT) tools is related to translation-oriented writing. And she explains small ways in which optimisations can help to reduce translation errors, lost time, additional costs and unnecessary queries.

It makes sense to write clear and comprehensible texts. This almost makes even more sense if the texts are going to be translated. After all, The more comprehensible the source text, the more comprehensible the translation result – and the simpler, faster and cheaper the entire process. Translation-oriented writing contributes significantly to this. We explain the essential aspects, show what to look out for and reveal why those using the source texts also benefit from it.