icon
background image

ISO 18587 - MT is on the way

Human intellect plus artificial intelligence

Artificial intelligence is on the advance and has been used in the field of translation for quite some time now in the form of machine translation (MT for short). To name a common example, decent translations can be produced in seconds using a smartphone and an online translator, making our lives easier. However, the solutions we rely on for simple daily tasks are not equally applicable at a professional level. In the field of corporate communications, marketing or specialist medical translations, the error rate typical for artificial intelligence is still far too high.

For specialist medical translations, machine translation output can not simply be used „out of the box.“ It must be professionally checked and edited by language experts. In this process, which is called post-editing, the post-editor proofreads the machine output, adapting and supplementing technical terminology to produce a translation that is equivalent to a purely human translation.

 

Translation process: infographic & individual steps

 

  1. Analysis of the source text for its suitability for machine translation:
    language combination, specialist field, style, format
     
  2. If the source text is suitable for machine translation, we will work with you to define the desired quality level for the upcoming post-editing (full post-editing or light post-editing).
    Full post-editing produces a result comparable to the translation by a qualified human translator.
    Light post-editing merely results in a comprehensible text, making it very suitable for internal communication, social media articles, e-mail correspondence, chat bots, etc.
     
  3. Machine translation
     
  4. Our post-editors will then process the machine output according to the quality level you selected, based on the ISO 18587 standard. If you already have a translation memory or a terminology database with us, these will also be taken into account during post-editing, depending on the selected level of post-editing. We would also be pleased to integrate provided reference materials.
     
  5. The final product is then subjected to an internal quality inspection.

 

Pros

  • shorter translation time
  • cost-savings
  • shortened time-to-market cycles: information products can be distributed multilingually considerably faster, even with very large volumes
  • no loss of quality compared to a human translation if full post-editing is applied
  • total flexibility due to freely selectable quality level (light or full post-editing)

 

Cons

  • not applicable to all types of text
  • fluctuations in quality due to disruptive factors in the source text such as inconsistent terminology, incomplete punctuation, incorrect spelling and complicated style of writing. MT requires discipline in the source text

 

  • if a basic set of terminology is already established or the machine has been trained
    (in particular segments from translation memories)
  • if time is more important than quality or a lower quality level is sufficient (light post-editing)
  • if the text is straightforward (ideally the author has already taken MT into account):
        - short sentences
        - little ambiguity
        - standardised terminology and syntax
  • if the text is used exclusively for information purposes
    (internal communication, social media posts, e-mail correspondence, chatbots; in combination with light post-editing)