Post-Editing (PEMT)

Post-editing of machine translation (PEMT) is another service that requires a specific, mindful approach. Machine Translation (MT) can greatly increase the speed and reduce the cost of the translation process; the trade-off is decreased accuracy. There are 4 standard types of MT: statistical, rule-based, hybrid and neural. Neural MT engines are fast becoming the most commonly used by LSPs, as once trained they produce the most accurate results.

However, machine-translated text is still nowhere near the quality level of human translation, because AI cannot replicate the full extent of human consideration, experience and nuance. This is why MT is only suited to certain text types and post-editing of machine translation is required.

I offer graded levels of post-editing, ranging from editing to enable gist understanding of a text (extraction of pertinent information) to a more polished translation that reads just as well as human translation. The key to deciding upon the service that you require is in a “good enough” approach – what is the end use of the document and who is the target audience?

Pricing and time frames for these projects are dependent upon the quality of the raw MT output; poorer quality MT will obviously require a great deal more processing (and usually full translation) than neural MT produced from a trained engine with client terminology and ICE matches applied. I will always check through the raw MT output file before accepting a project, to provide a reasonably accurate time scale and rate.