What is post-editing, really?
Post-editing (or MTPE, for those who like their acronyms) is when a human steps in to fix machine translation output and turn it into something that people can actually use. The person doing this is called a post-editor, though some days it feels more like “rescue worker”.
You’ll often hear post-editing mentioned alongside pre-editing, which is when the client or typically the translation agency tidies up the source text before feeding it to the machine.
It’s important to understand that post-editing is not the same as editing in the usual sense. Editing typically involves refining human-written text, which we in the translation world often call revision or review. Post-editing, however, is about taking raw machine output and making it acceptable, sometimes by simply making it understandable, and other times by transforming it into something that reads like a professional human translation.
How much post-editing you do depends on the brief. Light post-editing means fixing up obvious errors to make the text understandable. Full post-editing goes further, refining tone, style, and clarity until the end result could comfortably stand next to a human translation.
With machine translation improving rapidly, full post-editing is increasingly being offered as a practical alternative to translating from scratch. Most CAT tools now come with post-editing functions built in, and it’s becoming a standard part of many translators’ workflows.
A brief history of post-editing (and how we got here)
Machine translation made its first real-world appearances in the late 1970s, at places like the European Commission and the Pan-American Health Organisation. Soon after, companies like Caterpillar and General Motors began using it. By the 1980s, people realised that machine output on its own wasn’t quite ready for prime time, and post-editing started to develop as a way to bridge that quality gap.
In 1999, post-editing earned its own special interest group within the AMTA and EAMT. Then came the internet, browser-based MT tools like Google Translate, and a tidal wave of machine-generated content that someone had to clean up. That “someone” was, and still is, us, the translators.
As globalisation picked up speed, the sheer volume of content needing translation grew beyond what human translators (even with CAT tools and translation memories) could manage. This is where industry groups like TAUS started promoting MT and post-editing as essential tools for keeping up.
Why human translators still matter
Despite all the buzzwords and shiny AI breakthroughs, machine translation still lacks one crucial thing: understanding. Machines predict word sequences; they don’t grasp nuance, context, or cultural subtleties. Humans, however, can read between the lines, spot idiomatic expressions, and ensure that a message lands the way it’s meant to.
That said, when machine translation is paired with skilled post-editing, it can streamline the workflow. Adaptive MT systems that learn from user corrections can cut down on repetitive fixes and improve efficiency. At least, in theory. The key is still human oversight.
Light vs. full post-editing – not all post-editing is created equal
It wasn’t until 2017 that we finally got a formal post-editing standard (ISO 18587:2017). Before that, everyone was winging it to some extent. Today, we usually talk about two types of post-editing:
- Light post-editing: Minimal fixing to ensure the text is understandable.
- Full post-editing: A deeper level of intervention to make the text polished, stylistically sound, and ready for publication.
Light post-editing is often used for internal documents or content that’s needed urgently, while full post-editing is for anything that will be shared externally or published.
Productivity and efficiency – theory vs. reality
Post-editing is often sold on the promise of being faster than translating from scratch. Industry benchmarks suggest that light post-editing can quadruple productivity (1 000 words per hour versus 250), while full post-editing should at least double it. But here’s the catch: real-world results vary.
Efficiency gains depend on a whole mix of factors, including:
- The quality of the machine translation output
- How similar the source and target languages are
- The complexity of the text
- The post-editor’s skill and experience
Some studies claim time savings of up to 40%, but others report gains closer to 0-20%. And in cases where the MT output is particularly poor, it can take longer to fix than to start from scratch. So, while post-editing can improve productivity, it’s not a magic bullet.
Post-editing as a profession – still finding its feet
Even after several decades, post-editing remains somewhat of a grey area in the translation world. It overlaps with both translation and editing, but it’s not quite either. Many translators are hesitant about post-editing, partly because it often comes with significantly lower rates, even though it can require the same (or often much more) effort.
That said, I believe MTPE has its place as long as it’s used responsibly. I offer post-editing services for certain types of content, but not regulatory texts or other contexts where machine translation has no business being involved. For general content, non-technical texts, or internal documents, however, MTPE can be a practical solution provided the process is handled with care and expertise.