An Australian startup revealed its flagship product, an earpiece that can interpret 8 different languages in real-time, at a United Nations event in Switzerland on Friday.
Lingmo International, a startup based in West Gosford north of Sydney, launched its TranslateOne2One earpiece at the UN’s Artificial Intelligence for Good Summit in Geneva, revealing that IBM Watson machine learning technology had been used for its algorithms.
Traditionally, converting one language to another orally in real-time is called “interpreting” whereas the term “translation” is reserved for processing text across languages with some delay. Lingmo founder Danny May, however, describes his product as performing “translation in real-time”.
“It’s a fully independent translation earpiece. And what I mean by independent is that it doesn’t require any connectivity to your phone by Bluetooth or wi-fi. A lot of our competitors do,” he told Business Insider.
TranslateOne2One can handle English (US and UK), Japanese, French, Chinese, Italian, Spanish, German and Portuguese.
May started his working life as a fully qualified plumber, before he entered the solar power industry. It was during this phase of his life, four years ago, that he encountered trouble on a business trip to China.
“My passport was stolen literally 4 hours after I landed,” he said.
“I went to a police officer and spoke to him using a competitor’s translation service and said ‘hello, how are you’. But it translated that as ‘hello, I love you’.”
The incident inspired him to create his own language conversion technology, which has taken an arduous 4 years to develop, as May didn’t want a word-for-word translator like the ones commonly seen on search engines.
“We’re moving away from that. We’re taking up to a 30-second block, [IBM] Watson then takes it and puts it into a coherent sentence,” he said, adding that field tests indicated people speak for about 15 seconds at a time.
The oral, rather than text, input was also a challenge.
“We really focused on the speech recognition side – for the simple reason that if the speech isn’t picked up correctly, then the translation will come back incorrect.”
After the algorithm was developed, the company turned its attention to the earpiece hardware, which May said was the “easy bit” that took only 12 months. The combination of the software and hardware is now undergoing beta testing and final aesthetic and ergonomic fine-tuning.
The startup is aiming for the consumer market, with May telling Business Insider that the TranslateOne2One is expected to be on the market for $US179 ($237). Pre-sale orders are planned to start this week with first shipments scheduled for mid-July.
The interpreter is also downloadable as a smartphone app.
The final device has 3G mobile connectivity, meaning that it has machine learning capabilities when it’s connected to the internet. The quality of the interpretations will improve over time, according to May, especially if similar phrases and words are heard.
“On average we’re going for 85% accuracy. We’re moving away from word-for-word translations and taking [phrases] in context – and currently 85% accuracy is what we’re getting from doing field testing.”
Lingmo was bootstrapped for the first three years of existence, then in December it took in just under $250,000 from private angel investors. May told Business Insider that he was currently working on a second round of capital raising.
The company also had assistance from the IBM Global Entrepreneur programme, which it joined last October.
“It’s an absolute game-changer for the global translation market, offering significant potential for both commercial and social benefit, which we’ve only just begun to explore,” said IBM Watson Master Inventor Neil Sahota.