Chartered Institute
of Linguists

Promising perfection

By Jonathan Downie

Should interpreters adopt the marketing approaches of automated interpreting devices, asks Jonathan Downie.

“Imagine being able to snap your fingers and become fluent in twenty languages.” That may well be the dream. It is also the first line of an advert for a device that claims to be able to deliver “professional grade translation” by going “a step further” than professional human interpreters. Waverly Labs’ Ambassador claims all that and more, and whatever you might think of their particular brand of marketing schtick, they have been successful. The company has raised more than US$6 million in crowdfunding alone, and has managed to sell two product lines.

We might have our own feelings about the claims made by the producers of speech translation devices, but we can’t argue with their marketing and fundraising success. At a time when interpreters are dealing with the results of Covid-19 and the vagaries of government outsourcing, might it be time to learn the strategies used by these companies to see if we can adapt them for our own purposes?

Language barriers as a problem

The marketing messages used to sell machine interpreting can be split into three key ideas. The first is the simplest: language barriers are problems. From Andrew Ochoa of Waverly Labs bemoaning the fact that, without French, he couldn’t chat up a French woman he liked1 to Google vaunting the ability of their system to let you order a pizza in Italy,2 the idea that linguistic differences create barriers is the foundation for everything these companies claim.

This message might be written in terms of the challenges that people face in the real world, from closing business deals to finding their way on holiday, or it might take the form of connecting linguistic ability with sexual attractiveness. One company, Logbar, controversially advertised their ili device by sending out a man to see how many kisses he could get using it.3 However the message is communicated, it betrays the same logical and cultural failures.

Portraying languages as barriers supports the ideas that learning languages is hard and that all you need are the right words. Get a machine that can turn a sentence in English into one in Japanese or Spanish or French and the language barrier seems to disappear. However, anyone who has learned a language knows that words are just the beginning. Sure, it might be useful to have a device that helps you order a pizza in Italy, but learning Italian and getting to know the culture will tell you what might happen if you ask for a ham and pineapple pizza in a traditional Italian pizzeria. Even phrasebooks include little cultural notes. Surely, there has to be something better than seeing languages as barriers and pretending that knowing all the right words is enough.

A triumph of technology

The second message from machine interpreting makers is that overcoming language barriers requires a technical tour de force. Watch any video on machine interpreting, or read any blog post on the topic, and you will be introduced to smiling but hard-working engineers, complex diagrams and explanations of just how hard it was to crack the problem.

Take, for example, the less controversial video from Logbar in which the founder of the company explains the background of the ili device, including technical drawings and him having one-way conversations with lots of happy professional-looking people. The message is clear: this man is leading the way to building a technical marvel.

Waverly Labs take the same approach. In the video for their new Ambassador device, you can see close-up shots of the circuit board that is embedded inside, alongside explanations of the “microphone array” and, of course, promises that your speech is “uploaded to the cloud” to be translated using a “hybrid model”. With the benefit of being able to promote a sequel to their first Pilot Translating Earpiece device, they can both praise their previous technical efforts and make impressive noises about their newest device.

There are two important points to note here. Firstly, there is no doubt that any automated interpreting technology does, indeed, require a great deal of technical expertise. That much is undeniable, even if we might want to probe the meaning of some of the terms they use.

More importantly, the technical jargon in these adverts always serves a purpose. These companies know how to make it seem as if they have thought through any problem their users might have and so can explain that this microphone array, or that circuit board, solves this problem for them. This leads neatly into the third, more subtle message.

Speech translation is perfect

Every single word in every single advert for every single automated speech translation device is aimed at a single central claim. For a technology company to sell any devices, they need to convince their audience that the results from their products will fulfil their clients’ requirements.

In Waverly Labs’ latest advert, they make a great deal of their ability to deliver “professional standard translation” for “professional users”. In the demonstrations of the technology, the quality of output is technically flawless. Indeed, even their initial dream sequence, which asks viewers to imagine what it would be like to snap our fingers and be fluent in 20 languages, rests on the assumption that the technology can deliver that fluency.

Look at any advert for any machine interpreting solution and you will see the same claim. Yet whenever such claims are tested, they are shown to come up short. In 2018, the Chinese technology giant Tencent decided to let their automated speech translation system, Fanyijun, take over the interpreting at their Boao Forum for business and political leaders. To say that its results were a disappointment would be an understatement. From interpreting China’s “belt and road” initiative as “a road and a waistband” to spouting gibberish,4 its output left the local press nonplussed and the international translation press in a jovial, almost sarcastic, mood.5

With ongoing concerns around Covid-19, it is not at all impossible that similar speech translation systems might be tried in even more important situations. In fact, there were talks of doctors in one Tokyo hospital being equipped with portable speech translation devices for the Olympic Games before the event was postponed.6 So what do we do?

Doing our part

I doubt there is a single interpreter reading this who would agree with the claims of the machine interpreting companies. We can all find the holes in their logic, and we all want to argue that languages are engines for growth, or welcome markers of diversity, rather than barriers. Yet we have to face the fact that, for the most part, the marketing of automated machine interpreting companies has been more persuasive than our efforts. It is us who need to change our tactics, not them.

When I studied the marketing of machine interpreting devices for my book, Interpreters vs Machines, I noticed that their claims are persuasive because they are based on real problems faced by travellers, and contain a clear response to these problems. Compare this to the usual marketing of interpreting, which focuses on accuracy, impartiality and confidentiality, as if we don’t need to explain their importance to clients.

If we are to beat the peddlers of the devices that are supposed to replace us, we need to turn their own tactics against them. We too need to be so in tune with those buying our services that we talk about our work in terms that make sense to them. Our marketing has to be based on the problems our potential clients have and how we can solve them.

Instead of trumpeting our qualifications, we need to talk about how we help companies grow, or make medical appointments more efficient. Instead of calling ourselves accurate and impartial, we need to demonstrate the difference we make to trials or business meetings or conferences. And we need to do all this using the words and phrases our clients use. As more cash is thrown at speech translation, learning to create persuasive marketing messages about human interpreting is no longer an option; it’s a requirement.


Notes

1 www.youtube.com/watch?v=NjjQ5cH_YzI

2 Google (2017) ‘Google Pixel Buds – Wireless headphones that help you do more’; www.blog.google/products/pixel/pixel-buds

3 www.youtube.com/watch?v=g82tUyukLck

4 bit.ly/3bjKTdM

5 bit.ly/2xr5xKA

6 bit.ly/2UloKX7