AI is revolutionizing work in an astonishing range of industries, and it has already made a huge impact in translation. Thus, many businesses today are considering how they can leverage AI tools to save time and money on their translation projects.
If your business has frequent translation needs, it’s tempting to lower your costs by reducing or even eliminating the use of professional translation services. Why pay for what AI can do for cheap or free?
Before you go this route for your translation tasks, it’s wise to consider whether the risks are worth the potential gains.
The different approaches to translation today are based on employing translation software in different ways. The most trusted approach is often called “machine translation plus post-editing.” This method leverages the utility and power of translation software and combines it with oversight by human translators.
In a nutshell: a source text is run through a machine translation tool, and a human translator carefully checks the translation for errors and makes improvements.
For some types of texts, machine translation plus post-editing can save considerable time and labour. For others, the human translator/reviser has to make a large number of changes to ensure the quality of the finished product. With so much human work required, the time and cost savings are often negligible.
So, regarding the question of whether machine translation is faster and more affordable, the answer is: sometimes. But the “machine translation plus post-editing” method still requires the intervention of a human, and often involves as much time and expertise as before AI arrived on the scene.
So what happens if you simply eliminate the human translator and trust AI to handle your translations? How do we know that AI can’t handle such tasks on its own? Let’s take a closer look.
To get a better sense of the strengths and weaknesses of AI translation tools, let’s take a piece of text and run it through a popular and powerful translation software. We’ll use a fairly typical passage from Les Affaires magazine for this experiment.
The French original :
Groupe TVA a annoncé, la semaine dernière, qu’il mettait à pied 547 employés, soit 31% de ses effectifs dans un contexte difficile pour l’industrie télévisuelle québécoise.
Questionné sur le sujet, le grand patron de Québecor, Pierre Karl Péladeau, n’a pas voulu s’avancer sur les montants qui seront payés en indemnités de départ avant d’avoir terminé ses négociations avec la partie syndicale.
The machine translation :
Last week, TVA Group announced that it was laying off 547 employees, or 31% of its workforce, in a difficult context for the Quebec television industry.
Questioned on the subject, Quebecor boss Pierre Karl Péladeau would not say how much severance pay would be paid until he had completed negotiations with the unions.
Looks pretty good, right? Most seasoned translators would agree that it does; but not good enough.
“…in a difficult context for the Quebec television industry…”
Here we see a poor translation decision that anglophone translators are trained to catch immediately: the problem of automatically translating contexte (FR) as context (EN).
To clarify the issue, let’s consider a dictionary definition of “context” (the English word):
“the circumstances that form the setting for an event, statement, or idea, and in terms of which it can be fully understood.”
This definition seems to suggest that “context” works nicely in translating “dans un contexte difficile”. But the last part of the definition points to why the solution isn’t quite right.
In common English usage, context is about understanding (“and in terms of which it can be fully understood”). Thus, to anglophone readers, the translation is saying something like “in a difficult background for the Quebec television industry.”
The thought conveyed is slightly illogical, which makes the phrase awkward. A better and more idiomatic solution would be: “during/in a difficult time/climate/period/etc. for the Quebec television industry.”
Readers will still understand the idea behind the machine translation’s choice; but many will stumble on the word “context,” sensing that something is off. This poor word choice signifies a dubious decision, and every bad decision by a company–even a small, seemingly insignificant decision–reflects on its credibility.
“Quebecor boss Pierre Karl Péladeau…”
The issue here is pretty obvious, to anglophone readers at least. An article of this type would not use a colloquial word like “boss.” Instead, it would use head of Quebecor, Quebecor’s CEO or something similar.
Is “boss” incorrect? Not technically. PKP is, in fact, the “boss” of Quebecor. In the context of this article, however, (and by the way, the English word “context” fits perfectly here!), it’s wrong. The New York Times would never label Bill Gates “the former boss of Microsoft” unless their tone was tongue-in-cheek. A machine translation tool, however, might not be aware of that tonal distinction and where to employ it.
Is the ill-judged choice of “boss” here a big deal? After all, the reader will still understand the meaning. I would argue that yes, it really matters. As we saw earlier, every poor decision increases the reader’s distrust. If a business employs a writer or translator who makes unprofessional choices, why didn’t they select a better resource? What other imperfections are they willing to overlook, and what other corners are they willing to cut?
“until he had completed negotiations with the unions…”
This problem is a subtler one. There’s no serious mistake here; the translation works.
However, one of the goals for good professional translators (and this isn’t common knowledge outside of the industry) is to improve the source text, without altering the intention, meaning or general style.
This objective might sound surprising: shouldn’t a translator simply reproduce the text as closely as possible, warts and all? In fact, the “improve the text” approach pays off, time and again. Businesses want high-quality texts, flawless language and clearly expressed ideas. They want quality writing that reflects well on their identity and boosts their reputation. When machine translation starts with a poorly written and logically unsound source text, it produces a faithful translation, with all its blemishes. Skilled human translators, on the other hand, give you something better to publish.
So, back to our third example: what problem might a skilled human translator find in the translated phrase “…until he had finished negotiations”?
The wording in this phrase seems to focus all the action on KPP. In actual fact, however, negotations always take place between two (or more) parties; they’re a back and forth activity.
Therefore, a skilled translator might decide to shift the emphasis just slightly. A better solution? Something like “until negotiations with the unions were concluded” would work nicely.
In just two short paragraphs taken from a typical news article in Quebecois French, our translator found two significant problems and one phrase that could be easily improved. And this experiment was conducted using a software that, in our translator’s experience, generally produces more optimal results than others.
The takeaway: without oversight from a skilled professional translator who knows how to spot problems that amateurs will miss, machine translation can–and often will–produce errors, awkward choices and poor style. These are problems that impact the reputation and credibility of your business.
Despite everything we’ve heard, we aren’t at the point where machine translation tools can work autonomously. Human intervention by skilled and experienced professionals remains a crucial part of the process.
Wenovio is here to help you with that process: we have the expertise to optimize the most powerful translation tools available and ensure that every translation is of unsurpassable quality.
Découvrez comment ChatGPT peut transformer votre processus de rédaction de blogue avec cette approche structurée.
Découvrez les limites de ChatGPT dans le marketing numérique et apprenez comment maximiser son utilisation.
Un audit de site web mené avec rigueur vous permettra d’en identifier les lacunes et vous offrira des pistes concrètes d’améliorations.
Apprenez-en plus sur les critères d’évaluation pour évaluer les soumissions pour un site Web provenant d’agences spécialisées.