It’s 2024 and the AI hype is all around. In particular, it seems to be the consensus that AI now does absolutely super translations. Well, yes. Maybe it depends on what you mean exactly by ‘super’. AI-assisted texts are almost free of grammar errors and they also seem to flow in a certain way. Since these are qualities that take hard work if English isn’t your first language, I can see how you would find that attractive. But what if your text has more work to do than just be alright? What if you need to communicate a complex set of technical subject matter? What if you not only need to express the technical content accurately, but also make it so relevant and interesting to your readers that they willingly make the effort to read your story and understand it? Is it possible to achieve this kind of communication, and can we talk logically about how to do it? I believe we can, but to do that we need to get to grips with other aspects of text quality.

How much do you know about how a human translator works? I think one aspect of translating that might surprise many non-translators is just how many decisions are involved. On the one hand there are what we could call ‘strategic’ decisions – things I derive from my understanding of the readers and the communicative purpose of the text. And then there are many details on which I do some research and choose a particular option based on very small amounts of data or even a single data point. In both cases I am making decisions based on my knowledge and logical reasoning – a very different process than an AI reproducing patterns it has found in masses of texts. If you have never made a serious attempt at translating specialized texts, I think you might be surprised at how quickly we can find ourselves in a space where there is only a tiny number of relevant existing texts to draw on, and where even these are not completely reliable.

Let’s take a quick tour through some of the types of decisions a translator has to take. At least, these are things this translator often considers:

  • Whether to leave out or add information. Does that sound like an awful thing to do in a translation? But take this example: The German version introduces an English specialist term which it then explains in a way which would not be needed in English, because in English the term is self-explanatory. If the goal is to produce a highly readable translation, this needs to be left out. Or the other way round: a term whose meaning is obvious in German will need explanation in English (if it is something that only really exists in German-speaking countries, maybe the best solution is to define it and then keep using the German word in the English text; but there are other options too).
  • Sometimes the difference a translation has to bridge is not just between the two languages, but also between subject traditions which have developed independently in the different languages. Often in a specialist subject there are different conventional ways of defining and explaining its concepts, so that certain topics are discussed with a different logic. For example: structural engineering and wooden construction.
  • Whether there are regulatory environments in play; considering whether particular expressions could be understood as implicit references to particular standards or if this impression needs to be avoided.
  • Whether the beginning of the text needs reorganizing in order to make it work in the target language. German texts often take their time to get to the point. We don’t like that in English. Instead, we expect a first paragraph to give us an accurate announcement of what is coming in the rest of the text. And so it’s often necessary to sharpen up the beginning or even write a new one.
  • Just as the whole text needs a good first paragraph, in English we also expect the beginning of a paragraph to clearly indicate what the paragraph is about. We want a topic sentence. Good topic sentences make the whole text easier to navigate.
  • The translator reorganizes information in a paragraph in order to arrive at a progression of ideas that is suitable in the target language; for example, by putting events in their chronological order. Sometimes information and stories in German texts are arranged in a way that would seem jumbled in English; if I have the freedom, I might move some material between paragraphs so that each one tells a clean story about just one topic.
  • Whether to rearrange information inside sentences so that a sentence doesn’t contain too many contrasting or unrelated items of information.
  • Which pieces of information are new and important at each point in the text, and should they be given a more prominent position in terms of the sentence structures? Or even their own sentence? New information in German is often in a position inside the syntax of a sentence that would seem inappropriately ‘buried’ if we reproduce the same structure in English.
  • The topic-comment structure inside sentences, and sometimes rearranging the information so that the beginning of the sentence connects to previous information, or sets the scene for what’s coming in the rest of the sentence, and the most important/newest/most specific information comes right at the end.
  • Which of the grammatically possible solutions for a sentence, that put the information in the right order, is easiest to understand. Even among sentence variants that are grammatically correct, German and English have different preferences.
  • Whether German pronomial adverbs such (e.g. damit, deswegen) can be simply translated or need to be resolved into a more specific phrase, in order to make their linkage to previous or subsequent points in the text as clear in the translation as they are in the source text. Same applies to other references between different sentences, too.
  • Whether the best way to handle an idiomatic expression is to replace it with an equivalent idiom or an ad hoc phrase that isn’t an established idiom, or if it’s better just to leave it out and use plain language.
  • Whether specialist terms are introduced in a way that will make them understandable for readers.
  • Whether what seem like mistakes and contradictions in the text are real and which of them is important enough to ask the client about.
  • If the text is next to images: making sure the text plays well with the images. In some cases the text/image interplay is so important that I rewrite the text based on what I can see in the image.
  • Whether to adapt headings and headlines so that they work well and fit with the conventions of the target language (e.g. adjusting the explicitness).
  • When a simple translation of a word won’t be understood as intended and a different word or phrase or a completely different treatment of the sentence is indicated in order to communicate the same idea.
  • How to elicit and encourage feedback. Listening to it; explaining the reasoning behind certain translations, learning from the client’s subject-specific language knowledge and reworking the text accordingly.
  • When to ask for/suggest modifications of the source text to make both versions sit well next to each other (important in bilingual brochures, etc.)

So, this was not an attempt at a complete list; it’s just the things that occurred to me quickly. But hopefully you can see that a successful translation is a solution to a puzzle with many parts. It is an attempt to create a text that satisfies a lot of different preferences and purposes at the same time. I believe readers notice when someone has made a good effort to write a text exactly for them: to make it relevant, easy to read, easy to navigate and easy to understand. That kind of text stands out from all the junk. And it’s that level of work you can expect from me.