Machine Translation with Human touch: Translating Customer Support Interactions
As businesses look to expand beyond their home markets, they must automate a number of support processes in order to satisfy the demands of a growing and increasingly global customer base. A customer base that often expects great service, in their native language.
Machine translation is an important component of many enterprises’ automation strategy. It could make replying to tickets more efficient, improve customer experience, and even help companies expand to different countries without having to hire native agents.
And yet, most customer service leaders are skeptical about introducing this technology into their workflows — and with good reason.
For most people, Google Translate is the first thing that comes to mind when somebody mentions machine translation (MT). But would you really trust it to translate everything you need to tell your customers?
Probably not, considering mistakes like mistranslating the name of a Spanish food festival as “clitoris festival1,” or identifying the phrase “Ooga Booga Wooga” as Somali2.
Provide Multilingual Support Without Sounding like a Robot
Great talent can’t single-handedly power your support operations. Even if you have the best customer support agents in the world, the languages they speak will determine who they can help effectively. So, what are your options if you need to provide customer support in markets with different languages?
You can hire a crowd of native agents and train them, which is costly and time-consuming. Or you can automate translation, which will reduce your costs and also make your teams more efficient.
At first glance, the choice seems easy. Who wouldn’t want to reduce costs and improve efficiency?
Nonetheless, there’s one significant detail that prevents most customer service managers from automating translation, and that’s the quality of the outcome.
So how can you make it work? How can you drive quality translation without sounding like a robot?
Over the past few years, I’ve been working on artificial intelligence technologies, particularly in natural language processing, a subfield focused on the analysis and synthesis of natural language and speech. But it was only when I became Director of Applied AI at Unbabel — a company delivering enterprise-quality translations of customer support interactions — did I realize that machine translation (MT) will play a major role in the future of customer service.
How does Machine Translation Work?
First, we need to understand how MT works.
In the past, machine translation systems were trained to read parallel sentences, which was a bit like teaching a parrot how to talk. The parrot may be able to do it, but will not know what they are saying or responding to. They are just mimicking the sounds we teach them to make.
However, the latest advances in the field have made substantial improvements to this technology.
A good example of this is neural machine translation (NMT), which has rapidly become the new state-of-the-art technology. It essentially consists of having a computer system act more like a brain by imitating biological neural networks, progressively learning and improving with more data. So in order for NMT systems to work and improve over time, you need human translations to feed into the systems and train them. Once the system receives all the data, it starts to learn patterns and produces better translations.
While all of this has allowed us to take machine translation to a whole new level, it’s still not enough. Even these new NMT systems aren’t necessarily up to the job of delivering the consistent tone of voice and near-total accuracy that modern enterprises require.
So what are these machine translation systems missing? Human oversight. You just need to strike the perfect balance between human expertise and artificial intelligence.
The Perfect Combination: Machine + Human
Pure machine translation systems lack the ability to adapt to different support channels or domains such as fashion or sports. They are also unable to customize to the requirements of specific clients, and likewise are unable to choose the right tone of voice (i.e., formal vs. informal).
In short, pure machine translation systems lack the human understanding required for interpreting cultural references and contextual cues. Today, however, NMT combined with advanced, automated quality assurance and post-editing by humans, ensures that translations are clear, native-sounding, and often delivered within twenty minutes.
This is a game-changer for customer service where it’s really not just a matter of quality but also translation speed3. In a world where customers are not willing to wait for more than 10 minutes4 to get their problems solved, attending to their needs in their native language on time is crucial. And this is where AI-powered, human-refined translation can help. This is what we focused on when we built Unbabel for Freshdesk.
The integration brings the NMT translation with human editing straight into Freshdesk, allowing you to help your customers in their native languages. This means that you can expand worldwide or support your existing international customers better without disrupting your workflow or hiring agents that speak all the languages that you need.
Machine translation may not have reached maximum potential, but it has come a long way toward meeting critical business needs, and this is only the beginning.
1 – https://www.theguardian.com/world/2015/nov/03/google-translate-error-as-pontes-spain-clitoris-food-festival-grelo-galicia
2 – https://www.buzzfeed.com/ikrd/people-want-to-know-why-google-translate-thinks-ooga-booga?utm_term=.dhDqrkN3AN#.vm5A9BXG3X
3 – https://unbabel.com/translation-speed/
4 – https://www.interactions.com/resources/customer-experience/understanding-customer-effort/