Let’s begin by defining “post-editing”
Post-editing (Machine Translation Post-editing—MTPE) is a service that entails the correction/improvement by humans of machine-generated translation output to ensure it meets an acceptable level of quality.
Post-editing is NOT proofreading—the edited text may afterwards be proofread to ensure quality if that’s what the customer needs and requested.
The level of quality desired by the customer, and agreed upon in advance with the translator, is variable:
- Light post-editing consists of correcting the content to make it acceptable and understandable.
- Full post-editing is also making the content accurate and appropriate on a stylistic level.
Post-editing is used when unedited machine translation isn’t acceptable, but, given the project’s specificities, a high-quality human translation is not required.
This type of service can be two times (full editing) or even as much as four times (light editing) more productive than manual translation.
However, post-editing is not a novelty! It has been around for a long time.
Post-editing is a form of human-machine cooperation in translation that has been around ever since MT operating systems were invented (read more here).
It has been a common practice since the 1990s, although its use was not well regarded.
However, recently, there’s a growing interest in post-editing from the translation professionals’ community in general and from clients, largely due to the increasing quality of machine translation. And we cannot leave out the recent availability of reliable and free software, unlike in the past.
As a result, with processes in the translation industry changing dramatically, the practice has “left the closet.”
For approximately thirty years, the post-editing profession was not standardized and lacked established and widely accepted guidelines.
In 2017, the ISO 18578 2017 standard was finally published: Translation services—post-editing of machine translation output—Requirements.
Post-editing is closely related to translating and editing, but not all post-editors are translators and not all translators are post-editors!
There are those who say a person who is bilingual but has never worked in translation will be easier to train for a post-editing job. That’s arguable, of course!
Many current post-editors are in fact professional translators, working as in-house employees or freelancers.
But also, many professional translators and translation agencies still refuse to offer post-editing services, mostly because it’s paid at lower rates than standard translations.
There are customers already requesting professional translators to post-edit instead of translating from scratch! And not because their project requires only an understandable translation, but because they (mistakenly) believe that we will offer them the same quality at a much lower cost! This is not the case for a number of reasons.
Post-editing and technological progress
Yet, post-editing can be profitable for linguistic professionals.
With the rise of artificial intelligence and machine learning, it may even become a fundamental service in the future.
The general public assumes it’s easier for translators to work on a machine-translated text, whereas linguistic professionals expect a poor-quality translation and lots of work to review it. They believe machine-translated content requires too much work to be worth the pay.
In the past, it was often easier to work directly from the source text and translate from scratch than to post-edit the usually poorly generated content. But advances in machine translation, mostly driven by the post-edited text being fed back into the engines for learning, now allow for a higher-quality output, and this has been increasing the demand for this service.
In fact, if the output is provided by a neural machine translation engine, the quality of the content for post-editing is astonishingly higher than it was in the past, therefore, requiring substantially less post-editing effort.
Plus, the majority of computer assisted translation tools currently support post-editing of machine translated content.
In sum, with technological progress, full post-editing is becoming a valid and relevant alternative to manual translation.
But why is post-editing important if MT engines are getting so good at it?
Machine translation will never be comparable to a translation by a competent human professional.
A machine, no matter how many nuances it can learn and memorize and apply, will never have the sensitivity and insight of a human being. And these are key factors for a translation that doesn’t distort the original message.
A good post-editor will drastically improve the quality of your machine-translated content, ensuring accuracy and correctness that no machine can rival—at least for the time being!
We can all agree that one of the main reasons for a customer to resort to machine translation (MT) is saving time and money, right?
In actuality, the idea of saving time and money by using MT is often misinterpreted.
The quality of a human touch is what will allow entrepreneurs to streamline their global projects. For example, it can even help them learn about common MT errors and be able to handle them in future projects.
And how is machine translation post-editing priced?
The price should depend on the type of text and content domain, the quality of the MT output you need to work on, etc.
So, common sense tells us price should be based on the actual effort the post-editor puts into the process.
Some language service providers (LSP) do charge per-hour rates, calculating a given number of hours by dividing the project size by the average translator output. Of course, productivity and volume estimates will also depend on heterogenous and unpredictable factors that skew the calculations, such as different post-editing skills and error type (minor to severe) and quantity.
In fact, it’s not advisable to charge based on time or the number of edits. It could even give rise to a conflict of interest. A not-so-honest post-editor could correct more than necessary, making a large number of edits only to charge more.
Recent surveys indicate that the most common alternative to the hourly rate is a per-word rate.
If the customer only needs a light post-editing, and the professional is expected to revise the target content to make it acceptable, then, customers should expect about 35-40 percent off the regular word translation rate.
On the other hand, if the customer requests a complete overhaul of the source text that leaves no trace of machine translation, a full revision will sometimes require more effort to process than it would to translate from scratch.
So, we can say a full post-editing would be 60-85 percent effort, so the discount should be made accordingly, even if post-editing effort is difficult to predict.
All things considered, the answer to the initial question is yes! This is a relevant skill
And it’s a complementary competence that’s worth attaining if you don’t want to miss the train of progress.
If you have a translation that is a reasonable candidate for post-editing, be sure to contact Verbarium for more information!
If you still want to know more, take a look at our further reading suggestions here: