Ben Guttmann
The web is made for everybody. In an economic moment where more and more of the internet is either being shut down (Gfycat just announced they are going dark) or being put behind lock and key (both Twitter and Reddit have recently erupted in turmoil over different API restrictions), it’s worth reminding ourselves that this smaller, more restrictive world is not how things are supposed to be.
A major part of making the digital world accessible comes down to the base level: can everybody understand the words written on the screen? If our message doesn’t pass this test, nothing else matters.
Nearly 59% of websites are in English. The next most common languages are Russian, at 5.3%, Spanish at 4.3%, and French and German, both at 3.7%. While English is the most spoken language in the world, there are over a billion Mandarin speakers, 609 million Hindi speakers, and 559 million Spanish speakers – and billions more among the rest of the world’s several hundred languages.
We’re all better off when the world’s knowledge can be shared with all of the world. So it’s incumbent upon us to create paths that make that possible. Luckily, when it comes to digital experiences, this is a whole lot easier than it has ever been before
The fastest, cheapest, and easier way to support multilingual access to your website is some form of machine translation – though this strategy has actually gotten a bit harder to make use of than it once was.
You’ve likely seen machine translation in the form of a Google Translate widget, which was added to many websites over the last couple of decades. It was that little box in the header or footer that opened up to a big menu of language options, and clicking on one would instantly translate the page’s text using Google’s automated algorithms.
Google discontinued that widget in 2019 in favor of a more complicated to set up (but ultimately more elegant) solution, their Google Cloud Translation API. For those who want something simpler, there are a number of third-party plugins for WordPress and other platforms that make things easier. While many reviewers judge Google’s translations the most favorably, Microsoft and others also offer similar tools.
The benefits of having software do the heavy lifting are clear: huge amount of language access at a meager cost. Here in NYC, one of the most diverse places in the world, the government provides transit notices in just 9 languages and voter information in 15 – and both at great expense. But with Google Translate alone, you can immediately have your content available in 133 of them.
The problem is those translations are going to vary in quality quite dramatically. While machine learning and other advancements have greatly improved this automated translation, you’ll often end up with something usable but far from artful. This solution is built for scale, not mission-critical messages.
Nothing beats humans when communicating with other humans. Human translation is just what it sounds like: hire a professional translator to port your content into multiple languages and then make that content available on your site or app.
Translation services have become much more accessible with the rise of marketplaces like Fiverr, though they still will almost always be slower and more expensive than running your text through a piece of software. But in exchange, you’re far more likely to get stuff back that actually makes sense.
Once you have that content, you must enable the platform running your site or app to house and switch between the content sets. If you’re running WordPress, several plugins can do with just a few clicks.
However, because you’re manually entering and managing this multilingual content, you need to be constantly vigilant about two things:
There’s also a possible half-step here: maintain separate language versions of your site, but use machine translation via Google or even tools like ChatGPT to help maintain the content.
While human translation can produce higher-quality output, it is hard to scale that quality for sites with hundreds or thousands of pages.
But what if you didn’t have to translate every page?
Increasingly, I’ve found that the most elegant middle ground is a hybrid approach to translation. Instead of making every piece of content available in every language, distill your site’s core information down to a single landing page and then have that page translated by an expert into every necessary language. You can still make everything else available via machine translation, but this way, you can ensure that the most mission-critical content has had human eyes on it.
Granted, this works much better for a museum, with relatively static vital info, than it does for a newspaper with regularly updated sections and timely content. But most smaller businesses and organizations will benefit from this approach.
Not all languages work the same way. Some read left to right, but others read right to left or even top to bottom. Some use Roman characters, but plenty don’t (and many fonts only support one set of characters). And some languages are just longer or shorter than others: I remember we once translated an existing app into French, but French translations tend to run about 30-40% longer for the same content. We had to design all sorts of bubble gum and duct tape solutions to make everything fit.
Through all of this, though, there’s only one real solution at the scale the internet operates: machine translation. While the results aren’t always perfect, it’s the only route that can work on millions of instances for billions of users.
And the only way that it works at that scale, for the web at least, is to rely on the common point of access, the browser. This is part of why Google discontinued their widget, because now most browsers can process translation automatically without relying on the site’s designer or administrator. Chrome, Firefox, Edge, and Safari all offer translation tools built into their software.
This can all feel a little dry and technical. But as author Ken Liu wrote, “Every act of communication is a miracle of translation.” If we want to connect and communicate truthfully and effectively, we need to examine the infrastructure of how we pull off the miracle.
