I proofread documents in French or German that have already been translated by a third party and/or written by a client.

I am also increasingly carrying out post-editing assignments. This is because the neural machine translation tools used increasingly by professionals still have many limitations and are far from perfect, especially in the legal field.

The quality of machine translation often leaves much to be desired, frequently containing numerous inconsistencies, such as: 

- lack of consistency in terminology: for example, a single term in the source language is translated into several different terms in the target language;

- false meanings, etc. 

This can be explained, among other things, by the fact that neural machine translations use terms from all sources; for example, for a translation from French into German, the machine will search for legal terms in German using all available sources in the German language, even if they come from other German-speaking legal systems, such as the Austrian or Swiss system, or even sometimes German translations that have been made to translate terms from a different French-speaking system, for example Belgian or Luxembourgish, and which are therefore neither appropriate nor accurate. 

Consequently, when using machine translation, it is important to check the translation very carefully, with a great deal of patience and, if possible, several times. It is necessary to correct mistranslations and misinterpretations, remove all errors in legal terminology and check consistency in order to harmonise the text. It is also necessary to check grammar, spelling, punctuation and style to achieve an acceptable result.