With the recent launch of Spanish language integration for Google Home, Google is poised to gain even more ground against Amazon’s Alexa-powered devices. The addition of Spanish language functionality expands Google’s share in the U.S. digital assistant market and into Mexico and Spain. Alexa currently supports only three languages, including English, German, and Japanese, while Google Home is set to offer over 30 languages before 2019. As Alexa falls behind, many wonder why it has taken so long to make digital assistants multilingual.
Lost in translation
“No hay pedo” is Mexican slang that means “there’s no problem,” but it translates to English as “there’s no fart.” Doesn’t quite work, does it? Language encompasses so many nuances that are specific to one region and dialect. English, too, has many variations, depending on where you’re from. For a digital assistant, knowing and speaking a language is different from just translating.
“How do you address a friend versus an acquaintance? When is an appropriate time for dinner? What jokes are appropriate? What jokes are funny?” asked Kelly Davis, head of machine learning at Mozilla, in an email interview. “The list is endless and not only language-specific but culture-specific. Jokes that work in Spain may be offensive in Mexico.”
Accounting for these differences takes an immense amount of time and effort. Companies must perform extensive research to understand these nuances and render them into code. This means even the largest companies in the world, like Amazon, have to invest several years into localizing their assistants. Amazon has even made moves to crowdsource learning with Cleo, a language tool powered by Alexa that learns from its users.
“You will need thousands of hours of this data from tens of thousands of people, and this must be done for each language and for each accent of that language you want to recognize. This is painstaking work, and there are no silver bullets,” said Davis.
Although we’re finally seeing tech behemoths like Google expand their offerings to include a wider variety of languages, companies that create digital assistants will always prioritize majority languages and cash markets. So while Amazon is covering the markets it deems most profitable, it’s leaving out many Spanish-speaking countries around the world, along with a large portion of the U.S. population that speaks Spanish.
An example of one of Alexa’s major fails this summer is its inability to answer queries about the World Cup in one of the most common languages among soccer fans. There are over 20 Spanish language channels and networks offered in the U.S. alone, which means Amazon is missing out on household DVR integrations and common inquiries like “Alexa, ¿a qué hora es el partido de la Copa Mundial?” (Alexa, what time is the World Cup on?). Such integrations will give Google Home an advantage until Alexa devices support more languages.
But the potential impact of adding multi-language support to digital assistants stretches far beyond scheduling meetings, changing the channel, and controlling a smart home.
“Think about how speech recognition could be used by minority language speakers to enable more people to have access to technology and the services the internet can provide, even if they never learned how to read?” said Davis. “Regular market forces will not help reach these people.”
The world of digital assistants is expanding. These huge advances in technology are appearing in more households than ever, and with the added pressure of increased language offerings from Google Home, we can only hope Amazon will catch up by making it possible for us to say “Hola Alexa” soon.
Cassie Tolhurst is a writer who contributes to Geek Girl and Stanford College Puzzle.