Machine translators vs. human translators
« previous post | next post »
"Will AI make translators redundant?" By Rachel Melzi, Inquiry (3 Dec 2024)
The author is a freelance Italian to English translator of long standing, so she is well equipped to respond to the question she has raised. Having read through her article and the companion piece on AI in general (e.g., ChatGPT and other LLMs) in the German magazine Wildcat (featuring Cybertruck [10/21/23]) (the article is available in English translation [11/10/23]), I respond to the title question with a resounding "No!". My reasons for saying so will be given throughout this post, but particularly at the very end.
The author asks:
How good is AI translation?
Already in 2020, two thirds of professional translators used “Computer-assisted translation” or CAT (CSA Research, 2020). Whereas “machine translation” translates whole documents, and thus is meant to replace human translation, CAT supports it: the computer makes suggestions on how to translate words and phrases as the user proceeds through the original text. The software can also remind users how they have translated a particular word or phrase in the past, or can be trained in a specific technical language, for instance, by feeding it legal or medical texts. CAT software is currently based on Neural Machine Translation (NMT) models, which are trained through bilingual text data to recognise patterns across different languages. This differs from Large Language Models (LLM), such as ChatGTP, which are trained using a broader database of all kinds of text data from across the internet. As a result of their different databases, NMTs are more accurate at translation and LLMs are better at generating new text.
As NMT technologies have become more widely available through online machine translation services such as DeepL, publishers, universities and other translation clients increasingly use them to translate whole documents. They then expect translators to do “machine translation post-editing” (MTPE), cross-referencing the machine translation against the original for a fraction of the price of a normal translation. Of course, in many cases, the translator’s edit of the machine translation is then used to train the machine – known as “human in the loop” translation – apparently moving us closer to a moment in which the human is no longer needed.
Although NMT full text translations have become much more readable, they are still far from being convincingly written by an expert native speaker. At present DeepL even seems to find it hard to do some fairly basic things like cutting sentences in half or reordering them, something which is always necessary in Italian to English translation. This will no doubt improve over time as it begins to identify more complex unwritten grammatical rules within the patterns of the various languages. But for now, to create a text that sounds like it could have been written by a native speaker, a translator will have to change the vast majority of the machine translation, and so it would often be quicker for them to start from scratch, particularly if they are supported by CAT.
Furthermore, a human translator makes numerous creative decisions based on their understanding of the tone, feeling, and sense of the original text, which inevitably also includes deleting bits, rewriting others, and even adding new elements to the text. Of course, the computer could make “creative” decisions like these based on probability in a certain context, perhaps by combining NMT and LLM technologies. But the most probable answer will not always be the best answer, and there is only so much of the context that the computer can take into account without understanding the text. The better it becomes at mimicking a human translator the more decisions like this it will have to make. And the more decisions it makes, the more room there is for error. These errors could either be relatively minor stylistic errors, resulting in a text that feels different to the original, or more serious errors in meaning. And these errors are more likely to be overlooked precisely because the text sounds more convincingly like a native speaker.
For example, the term “il popolo” in Italian would normally be translated as “the people” in English, however, in an article on the workers’ movement the computer translates this Italian sentence “Qual é il motivo per cui il movimento operaio non può avere come soggettività di riferimento il popolo?” as “What is the reason why the workers’ movement cannot have the working class as its reference subjectivity?”. “Il popolo” becomes “the working class” because the computer is smart enough to register that the article is talking about the workers’ movement and the working class is usually around when we’re talking about the workers’ movement. However, this article was specifically about the distinction between “the people” and “the working class”, and so the computer has completely confused the argument. In this case, the problem is precisely the computer’s attempt to take context into account with its “intelligent” non-literal translation. Again, although computers will of course become better at identifying the specificities of a particular context, in order to completely avoid these kinds of mistakes they would have to stop working with probability and instead understand the text they are translating, something which the current technology can only dream of.
As far into the future as I can envisage, skilled human translators and advanced AI translators will collaborate on their work, with the humans calling the shots and signing off on the final products. But the humans will be grateful for the assistance of the machines, because the latter will drastically reduce the drudgery and repetitious boredom of the low-level parts of the human translator's job. A skillful human translator is not needed for the routine, humdrum, monotonous labor that makes up the majority of translation work. On the other hand, a machine translator will never be able to undertake the subtle, sensitive interpretation required to render the sensitive poetry-prose of an immortal work like Abraham Lincoln's "Gettysburg Address" that is beyond the capability of any machine translator, but a genius of a human could do it.
Computer assisted translation — that's the name of the daily game.
Selected readings
"DeepL Translator" (2/16/23) — DeepL bills itself as "The world's most accurate translator", and it is indeed quite good, but — as I have repeatedly stated — Google Translate keeps getting better and better; overall, I believe that it is the best online translator, one of the main reasons being that it is able to identify frequently recurring constructions and locutions that are irregular in terms of grammar or idiomatic usage and provide them with a felicitous, human produced, ready-made translation, rather than just follow the usual rules the other translators are trained on; in my estimation GT is able to do this because it has available such a vast amount of raw data that is fed to it every day by those who do user searches and attempt translations, some of whom directly or indirectly complain that the basic training does not always work adequately for refractory phrases, not to mention that it is incomparably up to date.
Selected readings
- "Google Translate is even better now" (9/27/16)
- "Google Translate is even better now, part 2" (5/12/22)
- "The elegance of Google Translate" (3/10/18)
- "The wonders of Google Translate" (9/22/17)
- "Don't blame Google Translate" (2/4/18)
- "Google is scary good" (7/31/17)
- "ChatGPT writes Haiku" (12/21/22)
- "Alexa down, ChatGPT up?" (12/8/22)
- "Detecting LLM-created essays" (12/20/22)
- "Why electronic machine translation services sometimes seem to fail" (1/29/17)
[Thanks to IA]
Gunnar H said,
January 7, 2025 @ 7:24 am
Of course there will always be human translators for some tasks. I doubt that Google Translate's version of the Iliad will ever be the standard edition.
But if in a few years a streaming service is going to bring 200 episodes of a Korean soap opera to international audiences, are they going to pay to have a human translate the subtitles into German, Italian, Polish, Dutch, etc., etc.?
For many tasks it is not a question of whether the machine translation is as accurate and appropriate as an expert human translation would be, but whether the cost savings justify a slight reduction in quality, from the point of the view of the office making the decision.
I work for a company that offers a website in four different languages, and is already relying mainly on machine translation to produce the different versions, with more or less human oversight depending on the content. (Contract terms need both strict scrutiny and localization by legal experts, for example.)
And let's not forget that human translators also make mistakes from lack of understanding. To stick with subtitles, when the Star Wars movies were re-released in cinemas in the late 1990s, the Norwegian subtitles rendered "light sabers" as lette sverd ("light-weight swords").
More recently, in The Banshees of Inisherin, a sign for an Irish "Public House" got the subtitle Samfunnshus ("Community Center"), and when Barry Keoghan's character responds to a woman rejecting his advances by rambling "Just thought I'd ask on the off chance, you know… like 'faint heart' and that," the Norwegian subtitles declared that he suffered from a cardiac condition.
Philip Taylor said,
January 7, 2025 @ 7:50 am
Truly amusing gaffes, Gunnar — thank you for reporting them !
Michael Watts said,
January 7, 2025 @ 2:34 pm
This isn't something we need to speculate over; the relevant data is in the past.
The Chinese drama Like a Flowing River (大江大河) used to be available on Viki. Viki's subtitling system is that a team of volunteers with high standards create high-quality subtitles for the film.
But the rights changed, and the drama, and its sequel Like a Flowing River 2, were moved to YouTube. They kept the high-quality subtitles from Viki, but since the sequel had never been on Viki, they provided machine-translated English subtitles.
And those subtitles are completely incomprehensible. To watch the show, I had to find illicit fansubs through Reddit. The loss in quality is therefore easy to assess – it's 100% – but that doesn't seem to have been a consideration.
Chester Draws said,
January 7, 2025 @ 2:45 pm
For what it is worth, I find Google Translate produces better English, but that DeepL is more accurate from French, Russian and Polish.
GeorgeW said,
January 7, 2025 @ 3:14 pm
"On the other hand, a machine translator will never be able to undertake the subtle, sensitive interpretation required to render the sensitive poetry-prose of an immortal work like Abraham Lincoln's "Gettysburg Address" that is beyond the capability of any machine translator, but a genius of a human could do it."
I wonder if a human translator could render the "sensitive poetry-prose" of the "Gettysburg Address," not to mention capturing the language of the time it was written.
Jonathan Smith said,
January 7, 2025 @ 7:25 pm
WRT Mandarin in particular, state of the art is such that for say generating translated subtitles either direction, it would take a skilled translator with lots of time on their hands to generate a significantly better product than what you would get via systems like ChatGPT. "Joining forces" is of course the way to go wrt to both quality and (more obviously) time.
Josh R said,
January 7, 2025 @ 7:30 pm
Here's the reality of translation work, from my perspective as a professional Japanese-English translator.
Translation is highly skilled work, requiring high levels of ability in both the source and target languages, as well as the skill of translation, itself. Accordingly, it has been expensive work, particularly in rare languages like Japanese. The quoted article is 763 words long. 10 years ago, the going rate for such content would have been $0.15, even $0.20 a word, so to translate even such a short work to Japanese would cost $115-$150.
Nobody wants to pay that. Even 10 years ago, some clients would try to get by with using Google Translate and cut-rate MTPE. These days fewer and fewer clients want to pay top dollar for a quality translation by a skilled professional. Nor do they want to translation in a reasonable timeframe for the work required. They want it as soon as possible.
Professor Mair's idealized vision of skilled human translators working alongside AI collaborators is already a pipe dream. The market price for translation work has crashed; the current rates are half of what they were 10 years ago. A reliance on MTPE, by itself, would be bad enough. But add to this "translation farms." These are "translation companies" that use MTPE performed by low-skilled part-time freelancers for fractions of the rates that professionals are able to command. Heck, they don't even need low-skilled freelancers; they're happy to just warm bodies that are native speakers of the target language to pick out the obvious mistakes. They don't care that their business model is killing the market for skilled translation, they make money on volume, not quality. And their freelancers don't care, but to them it's just a little side job they do to make some extra money, it's not their living.
One might think that the above applies to a lot of general translation, but surely the more "artistic" market, movies, TV shows, and games that are meant to be seen by a wide audience, have certain nuances of tone and meaning that need quality translation? Nah. They just want it cheap and fast. Again, an army of part-time low-level freelancers comes to their rescue. Turn-around times for subtitles are insane. Typically, the middle-man companies that farm out jobs from the major streaming companies want the subs for a 30-minute show in two days. Maybe three if they're feeling generous.
The future is, with prices bottoming out, no one is going to go into translation as a living. The skilled translators will get fewer and fewer as turn-around times get shorter and shorter and rates get lower and lower. But no one will need skilled translators, because the internet has made a huge pool of pliable amateurs available. And a lot of the translation will be de facto MTPE, even when not especially requested by the client, because the amateurs will turn to Google Translate, DeepL, or even, God help me, ChatGPT when they get over their head.
Josh R said,
January 7, 2025 @ 7:32 pm
I didn't think dollar signs would give the comment software fits, but for the record, in my post above that is intended to be 15-20 cents, and 115-150 dollars.