The Future of the Translation Sector: The Slow Demise of Post-Editing

I've been thinking a lot about the future of the translation sector lately. I've come up with a theory this week that I want to share.

I'll already warn you: first there's a sad part and then there's a positive part. Make sure you read till the end.

The Evolution of Technology

The time translators typed their translations in Word with a paper dictionary on the table are long gone. Professional translators use translation tools to make their work more consistent and efficient. CAT tools, that allow translators to save their translations into a translation memory they can later re-use, have been around for decades. Most translators have embraced this technology, because it offers the advantages of consistency and memory while allowing creativity to flourish in the new text.

The next big step was machine translation. After a few years of bad quality that made it not worth using, the quality improved with neural machine translation (NMT) and better data sets (e.g. DeepL). The enthusiasm among translators has been mixed and considered the advantages and disadvantages, that's no surprise:

• Advantages: greater work speed, sometimes good suggestions for difficult words or sentences which support the translation process.
• Disadvantages: the technology is built in such a way that translators tend to just read and correct what's suggested (this is post-editing). That leads to more translation mistakes and literal translations, so the translation quality is lower than translating from scratch. It takes good concentration and effort to stay critical when reviewing the suggested machine translation. Being human, everyone slips sometimes, even when they try their best. Second disadvantage: there is no more room for creativity if you don't get the time and money to translate from scratch. That means the output of machine translation will be mediocre at best. Third disadvantage: freelance translators (most translators in the sector, since agencies don't use in-house translation teams anymore) are requested to discount their translation rates when they get post-editing jobs. If the client has agreed with post-editing quality as opposed to translation quality, the increased speed of post-editing could indeed be a reason to use a different rate. But what we notice in practice is that clients think they will get human translation quality, the agency asks the freelancer to post-edit and then adds the non-sensical remark "we expect human translation quality" (but we don't pay you for it).

The most recent step is LLMs (large language models), such as ChatGPT. I've recently heard from a freelancer who has integrated ChatGPT into his CAT tool, because he expected the quality to be better than when using machine translation. I can imagine there are companies who are already trying to feed text in ChatGPT and asking it to translate it into a different language.

At the moment, ChatGPT is known to make mistakes, it "hallucinates" because if it doesn't have enough data on a certain topic to give the answer that's most likely the correct one, it simply gives an answer that's off the mark.

We see something similar with machine translation: if the engine doesn't find the correct combination of words in its data sets, it just omits part of the sentence or adds something that isn't supposed to be there.

But how will these tools do 5 years or 10 years from now? They will get better. Where does that leave the translator?

Why I Think Post-Editing As a Service Is a Dead-End

There are 2 trends in the translation sector that we've seen for years:

1. Translation agencies (especially the big, all-round agencies) are putting up a fierce competition battle that drives down prices. The advantages may be that clients get cheaper services at short turnaround times and agencies are pushed to keep innovating. The disadvantages are that the quality of the delivered texts goes down and that freelance translators are paid less and less. They are forced to work faster with the use of tools and are still expected to deliver the same quality. This is an illusion.
2. The use of technology increases and the technology gets better over time. The advantages are that translators need to spend less time on correcting the text since the tool makes fewer mistakes. The quality will still not stand out (the average of a data set is per definition average), but at least the text should be correct. The disadvantages are that translators will do less and less work on the generated text, which diminishes their added value, and more texts than ever will be produced and automatically generated. High-quality or original texts will become rare (this is of course also an opportunity to stand out: quality will be admired).

It's in the combination of these 2 trends where I see the demise of post-editing: the added value of simple post-editing will dwindle down so dramatically, that for the texts with no special requirements, the added value becomes zero. Once that becomes clear to companies, they can just run their text through ChatGPT themselves and have the translation immediately and for free.

I expect the process to go as follows:

1. At the moment, companies are promised a translation made with "the most advanced tools for a cheap price".
2. Technology improves, companies use ChatGPT for other purposes and think about using it for translation: agencies will argue that they shouldn't take the risk and they can do it for an even cheaper price. Once that happens, the freelancers get paid even less and have to do more words per hour, which means they will change less in the generated translation (which reads pretty well at that point anyway).
3. One day, the company compares the received translation with the output of ChatGPT and notices that only a comma has been changed. The company calls the translation agency and asks "WTF are we paying you for?"

The Future of the Translation Sector: Added Value

All is not bleak. For those who hold on to post-editing as the future, they will have to fight an uphill battle against the uptake of technology in businesses if my theory is correct. Their job will be to convince companies that reading through generated output is necessary. They might even call it "quality".

My definition of quality is different: it's a text that stands out in all the right ways. This can be the case in different types of texts:

• The formulation of legal clauses to protect the company from being sued.
• Precise instructions to work with dangerous machines, so as to avoid workers getting hurt or even killed.
• Entering the market of a country with a different culture: marketing materials shouldn't be translated, but adapted to the culture of the new target audience.
• Authenticity and storytelling in marketing: how to make a text engaging and really resonate with the reader?
• Using the company's tone of voice throughout different communications.
• Translating texts that require expert knowledge in a niche.
• Etc.

I think it's possible that a part of the translation market will disappear. All texts where just making the content understandable will be translated by tools. The part of the translation market that will continue to thrive is the part that adds value.

So we all have to ask ourselves that question:

• As a company: what is the function of this text and which value do I expect of the translation?
• As an agency or a freelancer: which added value do I offer, apart from what the technology will be able to do as it further improves?

Word Atlas is an agency that offers translation and editing to the public sector and a few B2B sectors, mainly legal and financial companies.

We use tools only when they support us in delivering better quality or in delivering the same quality in less time.

We add value by focusing on multi-annual partnerships that allow us to develop a tailor-made offer to the company, focused on the company's needs and workflows. We love to learn the company's communication style and make it our own. If you are looking for a trusted partner, we could discuss your needs and see if we are the right match. Contact Word Atlas at [email protected]