#Open journalism No news is bad news

Your contributions will help us continue to deliver the stories that are important to you

Support The Journal
Dublin: 17°C Thursday 16 September 2021

Never struggle for directions to 'la playa' again: The future of speaking other languages

From multilingual earbuds to neural machine learning, barriers between languages are being eroded, writes Sarah Harford.

The way we live is changing fast. Every fortnight in our Future Focus series, brought to you by Volkswagen, we’ll look at how one aspect of everyday life could change in the future. This week: translating on your holidays.

‘OÚ EST LA bibliothèque?’ ‘Cá bhfuil mo dhenims dubha agus m’Umbro top?’

We’ve all studied other languages at some stage – whether we want to be able to communicate fluently in another tongue, or just know the basics for when we’re on holidays.

But technology has been gradually making its way into the classroom, changing how we learn in the 21st century. When it comes to picking up or perfecting a different language, new tech means that there are more tools at your disposal than the phrase book you may have brought overseas in the past.

Now if you’re looking to get a ‘cúpla focal’ or learn how to ask for directions to ‘la playa’, you can easily find videos and tutorials online, or sign up to subscription-based e-learning tools such as Babbel. You could also download the Duolingo app, which now has over 300 million users learning languages from English and Spanish, to Irish, Welsh and even Klingon.

There are also more novel methods being developed, such as ways to learn languages while watching Netflix, or using immersive technologies such as augmented and virtual reality to put users in ‘real world’ language scenarios.

High-tech translation

But with new developments in technology, will we need to continue learning other languages at all in the future or will there be devices that take care of translation for us?

You might be reading this article in English, but with a few clicks you could quickly turn the text into Swedish or German. Google Translate now translates more than 100 billion words every day, for free.

Andy Way, professor in computing at DCU who has a background in languages, explains that this is made possible by machine translation.

“The great thing nowadays is that everyone has used something like Google Translate, from my 12-year-old daughter to my 80-year-old mother. So people have some idea of what machine translation is – even if they might not understand how it works,” he says.

Researchers have been working on the concept of machine translation for more than 50 years, developing statistical and rule-based models for computers to try translate text into different languages.

Source: Shutterstock/Eiko Tsuchiya

However in the last few years a new system has emerged – neural machine translation – which is advancing translation technology. Basically, the neural system maps one language onto another using existing data taken from the internet or from institutions like the European Parliament that have to translate documents into multiple languages.

“Neural models use machine learning. So what we need is a large collection of parallel data – the larger the better – so the system can make sense of it and try to hypothesise words and phrases in your target language,” Way says.

“It’s all done with maths now but someone like me with a language background can bring in linguistic information and can see if the translation is good or bad.”

Joss Moorkens, assistant professor at DCU’s School of Applied Languages, adds that the introduction of neural models has made machine translation more reliable overall.

“We’ve seen that the quality has improved. The likes of Google Translate is far more fluent than it might have been four years ago,” he says.

“Even though the idea of neural machine translation has been floating around for a long time, it requires vast amounts of data for training, so it’s really only available for major languages like English, French, Spanish, German. It doesn’t work as well for languages where there isn’t a huge amount of data, like Irish.”

New hardware

The Hitchhiker’s Guide to the Galaxy by Douglas Adams features a handy universal translation tool – the babel fish. In the book, the creature is placed inside the ear and performs instant translations that can help the wearer understand any language.

But as machine translation systems improve, that science fiction may become more of a reality. In 2017, Google launched Pixel Buds – Bluetooth earbuds that aim to enable real-time translations between 40 different languages by connecting to Google Translate on the user’s smartphone.

The idea is that when you tap the buds and say, ‘Help me speak Japanese’, for example, it opens the Translate app on your phone. Then when you speak a sentence, it translates the words, transcribes them into your target language, and reads them back into your ear.

Bringing books like this on your holidays is already a thing of the past Source: Shutterstock

But Google isn’t the only player focused on this area. New York tech startup Waverly Labs is developing a similar earpiece using a cloud-based translation system, while Australian company Lingmo is working on a range of devices, including a smartwatch that does translations independent of your phone, and a translation earpiece that works without an internet connection.

To keep conversations flowing, Skype has developed a system that can be used for instant translations during video calls. Using a combination of machine translation, automatic speech recognition and text-to-speech technology, Skype Translator can currently translate conversations into 10 languages in real time.

“Despite the fact that it’s trying to string three imperfect technologies together, it works reasonably well, with only a five to 10 second delay,” Moorkens says.

“Elsewhere, there are a few more applications of those sort of speech recognition, machine translation and text-to-speech technologies that are being developed in Japan for the Olympics there in 2020.”

Although all these applications are still in development and none of them are perfect yet, Way says that machine translation is still reasonably good in most circumstances, and will continue to improve as more data becomes available.

“For me, bad English is better than perfect Russian because I don’t speak any Russian. So if you want to facilitate communication with someone and you don’t have a mutual language, then machine translation or some translation technology is the only game in town,” he adds.

The limits of machines

But while there have been major advancements in recent years, there are still some significant limitations when it comes to this kind of technology.

“Because of the large amounts of data required, once you start to get into a specific subject field, you might find that machine translation isn’t ideal for your needs and wouldn’t have the appropriate terminology,” Moorkens says.

“If you’re translating tweets or a TripAdvisor review, it might be a bit embarrassing if it’s wrong but it won’t cause any real difficulty. However if it’s something medical, if you’re trying to talk to a doctor, a mistranslation of a body part could be fairly risky.”

Mistranslation aside, another big problem with machine translation is ambiguity. Language has context, nuance, tone and cultural differences that can’t always be detected by a computer. As a result, Moorkens says that we’ll still need language experts and translators for the foreseeable future.

“There are improvements to be made to the technology, constant tweaks to the architecture. But in the short to medium-term, the idea that language learning won’t be necessary seems unrealistic to me,” he adds.

“Being multilingual is an incredible skill and should be considered as such. Translators’ jobs might change in the future and might involve curating some of the data that’s used for machine translation or working with the technology in some way, but there’s still a definite need for more translators.”

So although technology may make it easier to learn new languages and communicate with others in the future, it seems that we shouldn’t be tossing away our French and Spanish books any time soon.

“Will machine translation ever be as good as the very best humans? For me the answer is no because language is complicated,” Way adds.

“Machine translation is being used for all sorts of tasks now and it will be used for tasks that you haven’t even thought about yet in the next 10, 20, 30 years.”

The future is nearer than you think. Discover Volkswagen’s range of current and next-gen electric vehicles at volkswagen.ie/electric.  

Sponsored by:


Read next:


This is YOUR comments community. Stay civil, stay constructive, stay on topic. Please familiarise yourself with our comments policy here before taking part.
write a comment

    Leave a commentcancel