“Change is the law of life. And those who look only to the past and the present will surely miss the future.”
John F. Kennedy
Is the price of machine translation post-editing (MTPE) worth the effort?
As we’ve seen in the previous article, when MT produces quality results, it can actually increase productivity in terms of translation speed, allowing professionals to get a fair value for their effort. When MT produces poor-quality results, it can be a very unproductive task because the professional cannot get fair payment for all the work involved in post-editing.
So, what is the fair price? 🤑
Time does not stand still and we know nothing stays the same, so we can’t deny that the so-called 4th Industrial Revolution is upon us like a giant wave that we won’t be able to stop and from which we cannot escape.
Unlike MT, MTPE is a relatively new player in the language services market, which will hopefully be used to reinforce the translator’s work, rather than replace people—that would be the end of quality translations, and the beginning of major communication problems, with implications in personal, political, commercial, and also international relations….
As a relatively new and misunderstood language service, there is still no specific guide to pricing, and many translation professionals still refuse to accept MTPE requests.
Whether MTPE should be paid per word or per hour still generates debate, with some people preferring a word count and others an hourly rate. A word count rate should be 70–80% of a normal translation with any new words paid at the normal translation rate.
Let’s not be fooled by AI’s ability to produce bad texts that apparently look good!
After all, since it depends on the quality of the MT production, the time invested can be difficult to predict, so an hourly rate is definitely the fairest. ⏲️
The ideal price would be somewhere between the price of a normal translation and that of proofreading. Some agencies are willing to pay a fair rate, but many just want to save some money by pre-translating automatically and expect the post-editor to charge the price of a normal review.
However, the process of editing or reviewing a human translation is very different from the method used in editing machine translated content—even if the quality of the machine translation is higher than average, it will always require more work and time, as it will be necessary to carry out all the steps that a human translator would normally have already covered, such as carrying out all the terminology research.
Agencies expect you to always do a light post-edit (for an equally “light” price) but hope to get a high-quality end result.
And there are others who expect to pay 40–50% of the normal translation rate because they think it will take the professional half the time or less. This couldn’t be any further from the truth! 🙅🏼♀️
In the case of financial or legal texts, rather than a mere typing aid, machine translation can really speed up the work because it requires little creativity, as the original text is usually concise, accurate and readable. Because it does help get the job done quickly, a rate of 60% to 80% of your normal translation rate will, in most cases, become as cost-effective as a normal translation.
The problem is when the source text is not so simple, either because it requires creativity or because the sentences are too long for the AI to produce a readable text—and this can also occur with financial or legal texts.
While MT in the domains where it’s appropriate can indeed increase productivity, it may still require heavy human editing, if only to achieve terminological consistency in technical texts.
Given the circumstances, we’ve already seen that MTPE can take much longer than expected, but it also requires even more attention than that needed to review a human work, since we cannot rely on a minimum level of judgment and common sense.
Without a keen eye, we can easily be led to believe at first glance that a sentence makes perfect sense, when in fact its structure is awkward, or it has a glaring translation error that wasn’t easy to detect in a light reading.
It may even take several readings to filter out all the things that don’t make that much sense, which will take much longer than it would have taken to translate the text from scratch.
And, most likely, the customer will not want to pay the normal translation rate. After all, they already paid for the machine translation tool to avoid paying for translations in full.
By agreeing to review a poor-quality machine translation where they cannot—for budget reasons—devote all the time it requires, the post-editor risks taking the blame for errors in the final text and even receiving poor feedback.
Still, with productivity and cost savings as the main argument, the main market players have been pushing for MTPE to be dominant in the language industry, ignoring the unquantifiable factors such as the highly variable and unpredictable effort required of the translator, and the negative financial impact on language service professionals, the bottom level of the pyramid….
In sum, we agree that MT can be a useful and time-saving tool under the right conditions and if employed sensibly and skillfully to facilitate human translation work.
And let’s not forget that having the help of MT when translating or editing more or less raw machine translated content, does NOT mean that the professional translator or the professional proofreader needs less skills, competence and experience—or dedication. And it certainly is not an effective translation tool without human intervention.
It will always be a team effort, and never the victory of machine over humans. A less obvious misunderstanding translated into several languages at the speed of light is just that: a mistake.
Verbarium provides the professional MTPE services you are looking for. Do not hesitate to contact us to discuss the conditions of this service.
If you haven’t read the first part of this article, click here.