Ian Diggance Übersetzungen - Ihr Fachpartner für technische Übersetzungen Deutsch-Englisch

Machine translation post-editing

AI translations – 7 serious drawbacks

Machine translation post-editing (MTPE) is the process of reviewing and improving machine-translated text to make sure that it is 100% accurate and reads fluently. This is done by professional linguists who review the output and translation memory matches to ensure the highest possible quality.

Generally speaking, there are 2 core types of MTPE:

  • Light machine translation post-editing
  • Full machine translation post-editing

The 3 primary factors to consider when choosing the right MTPE method for the respective case are quality, time and cost. It is therefore essential to choose the appropriate post-editing strategy taking these aspects into consideration.

Light post-editing (LPE)

With light post-editing, the initial, raw machine translation (MT) output is only modified where absolutely necessary to ensure that it is legible and accurately conveys the meaning of the source content.

The post-editor should be especially mindful of errors that may be detrimental to the document’s purpose. Without a review, raw MT can create disastrous results. Editing should be kept to a minimum wherever possible. This approach can produce fast and cost-effective results, whereby a diminished quality must inevitably be accepted.

Full post-editing (FPE)

With full post-editing, the initial MT output is meticulously reviewed and amended to fully eliminate any errors. Where LPE focuses on the bare essentials of accuracy and legibility, FPE considers a number of factors such as:

  • consistent usage of the proper terminology
  • stylistic consistency throughout the document (and with other appropriate documents)
  • elimination of all grammatical errors

Fully post-edited machine translation should convince the target audience that it was originally created in the target language. This method is slower and more costly than LPE, but it does result in high-quality output.

Which MTPE method should be used?

LPE and FPE are to be considered as being a spectrum rather than a black & white choice. This is why MTPE requirements should be tailored to the translation project in question.

The post-editing priorities should be based on the aspects quality, time and cost. 

Post-editing can also be bypassed altogether. This strategy can be used for projects where the consequences of imperfect translations are negligible (e.g. internal documentation).

Using a machine translation tool for professional translations

When resorting to an MT tool, it is important to stick to scenarios where it can be used to the greatest effect without compromising on quality. 

These include:

  • Low-visibility or low-traffic content, such as internal memos, emails, social media posts for sentiment analysis, etc.
  • Low-risk content that does not have to be absolutely accurate or follow complex regulations. A good example here is user-generated content like product reviews where readers do not usually have high quality expectations.
  • Short-lived content, like chat or email support messages, customer enquiries, etc.
  • Large content volumes with a short turn-around, such as hundreds of product descriptions that need to be published quickly.
  • Frequently amended content, like feature and information updates.

Most of these cases will require light machine translation post-editing to assure accuracy and clarity. It may even be feasible to use the raw output if the content is not mission-critical.

An MT tool is only as good as how it is used

MT tools have obviously undergone a huge transformation since their humble beginnings (as practicable solutions) around 20 years ago. Today they can be used to rapidly generate high-quality translations for many (but certainly not all) purposes.

However, to exploit their true potential, they need to be used for the right kind of content and with differing, case-related levels of post-editing.

7 serious drawbacks with machine translations

It all depends on your quality expectations

If the quality of your translations plays a subordinate role because your texts are to be used in one or more of the scenarios described above, then you can safely resort to light post-edited machine translations.

However, if you are a manufacturer of high-quality machines, plants or products that involve extensive documentation or product descriptions with in-depth information, or you are an internationally operating service provider in a demanding sector, then you can simply not afford to settle for inferior-quality translations for your public image, products or business activities.

High-quality machines, plants and services also require high-quality documentation, which in turn needs to be professionally translated if it is to be of an equally high quality. If the documentation is pre-translated as part of an MT process, it will inevitably need to be fully post-edited in order to do justice to the high quality of your products or services.

This should be your aspiration as a leading manufacturer or service provider.

Prices for MTPE projects

In the case of machine translations that you have prepared internally and which you would like to be post-edited by us, special prices apply dependent upon the quality of the submitted machine translation and the intended purpose of the final translation. The tool used to create the MT (customer’s translation memory, DeepL, Google Translate, etc.) also plays an important role due to the varying quality of the output. Customer TMs usually deliver the best results. In such cases, the conditions only apply for the respective project.

The effort required to post-edit a machine translation should not be underestimated! Depending upon its quality and the required quality of the final translation, in some cases it could be cheaper and faster to have the content translated from the outset by a professional translator, especially if it contains customer-specific terminology, for example.