Skype's real-time language translator: Can it really connect the world?

Skype just made it easier to connect with people halfway around the world, even if they speak different languages.

The voice and chat application has expanded its real-time translation tool to any call to a mobile phone or landline in a user’s contact list, the Windows-owned company announced in a blog post Thursday. The tool is available through Skype Preview (a beta version of Skype) to members of the free Windows Insider Program.

Skype and its parent company are contestants in a machine translation arms race to help users lose their pocket dictionaries and tap into more foreign markets. Now that Google Translate can understand the text of 103 languages, the tech world has turned its attention to real-time voice translation. While machine translation is far from perfect, observers say its implications for busting through language barriers are huge.

"In theory, Skype Translation could be transformative," wrote John Pavlus for MIT Technology Review in 2015, months after the tool was first released just for Skype-to-Skype calls. "It’s like a version of the discreet live translation that world leaders enjoy when visiting the United Nations."

Mr. Pavlus acknowledged that the Skype translator has some of the same limitations as Apple’s Siri or Microsoft’s Cortana digital assistants, which both tend to speak over conversations.

"[But] even such crude automated translation is fairly remarkable," added Pavlus. "It is notoriously difficult for machines to recognize words and phrases quickly and accurately, and Skype Translator achieves a high level of accuracy using a technique known as deep learning."

The newest version of Skype Translator offers real-time translations for any call to other Skype accounts, mobile phones, and landlines, so long as they are in a user’s contact list. The tool is available for nine languages: English, Spanish, French, German, Chinese Mandarin, Italian, Brazilian Portuguese, Arabic, and Russian, according to Skype.

The feature isn’t perfect. Ever since Skype released a preview of it in 2014, reviewers documented its blunders. In October 2015, Nikhil Sonnad of Quartz recorded and dissected a translated conversation between himself speaking English, and another reporter, Zheping Huang, speaking Mandarin Chinese,. In the video, Mr. Sonnad and Mr. Huang are able to talk (via translation) using simple dialogue. But the Skype Translator tool struggles once they start to speak normally.

"It was unable to keep up with Mandarin spoken at a normal pace, sometimes not recording the speech at all and other times coming up with incomprehensible translations," writes Sonnad, "so that means one half of the conversation is largely missing."

But Skype Translator is always improving. That’s because it and other software like it work through deep machine learning, which is information processing loosely modeled on the way a human brain functions. In essence, it learns by both doing and making mistakes, as Alec Ross, a technology policy expert, and author of "The Industries of the Future," wrote in an essay for The Wall Street Journal in January.

"Whenever the machine translations get it wrong, users can flag the error – and that data, too, will be incorporated into future attempts. It is just a matter of more data, more computing power and better software," wrote Mr. Ross. "These will come with the passage of time and will fill in the communication gaps in areas including pronunciation and interpreting a spoken response."

These improvements will lead to a new iteration of globalization, argues Ross. Globalization was propelled by the expansion of English as the "lingua franca for business," he writes. When Korean-speaking and Mandarin-speaking executives meet at a conference in Brazil, for instance, they converse in English, not their native languages.

"The economic benefits of this new technology should be obvious. Machine translation will take markets that are now viewed as being too difficult to navigate and open them up," Ross wrote, citing the 6,000 inhabited islands and 700 languages spoken in Indonesia.

Silicon Valley appears to be acutely aware of this. Among the firms competing with Microsoft are Google and Facebook. When Google Translate, the machine translation service, expanded to 103 languages, its coverage extended to 99 percent of the online population, according to the search engine’s own estimates. When Instagram began to offer automatic translation on its photo-sharing service, it was able to translate 24 languages, while its parent company Facebook could translate 70.  

The translation services industry could also reap major profits for Silicon Valley, in addition to helping these tech firms reach new users. The industry was worth $40 billion in revenue in 2016, according to estimates from the consulting group Common Sense Advisory.

There are critics who see limits to machine translation. David Arbesú, a Spanish professor at the University of South Florida, detailed those limitations in an article he wrote in response to Ross’s. Dr. Arbesú argues machines will not be able to understand and translate the metaphorical language that is a foundation of human communication.

"Translating is an altogether different task than finding the nearest Starbucks, because machines aim for perfection and rationality, while languages – and humans – are always imperfect and irrational," writes Arbesú, in an article for The Conversation. "This is the paradox of computers and languages."

Ross also acknowledges machine translation still falls short in "accuracy, functionality and delivery." But he doesn’t believe that will be the case for long.

"A decade from now, I would predict, everyone reading this article will be able to converse in dozens of foreign languages, eliminating the very concept of a language barrier," he writes.