One of the cleverest plot devices used by Gene Roddenberry in Star Trek was the Universal Translator. It solved the problem of how multi-lingual humans could communicate with alien species. Now, an Australian plumber has teamed with IBM to bring the device to reality. (Well, for Earth-bound speakers, at least.)
Like many international travellers, Danny May found himself in an embarrassing situation. Speaking to a police officer in China, his smartphone translation app let him down. Instead of translating “Hello, how are you” correctly May ended up saying “Hello I love you”.
“The officer really didn’t like that very much,” said May.
May was in China designing and manufacturing solar products. Although he completed an apprenticeship as a plumber, May liked pulling things apart and seeing how they worked. Over the years, he became something of a technical whiz and built websites and dabbled in writing software. That expanded to the point where he partnered with developers in Pakistan who are still valued partners – so much so that his company, Lingmo, has an office there.
Thinking about how many other people were faced with similar situations whether they were travelling or dealing with locals for whom English is not a primary language, he saw an opportunity.
The company May founded, Lingmo, created a device called the TranslateOne2One. Although it looks like a Bluetooth headset, albeit a bit larger, the device operates completely independently to a smartphone. It has a cellular radio built in and requires a SIM card. It works on 2G, 3G and 4G networks.
When the TranslateOne2One’s microphone picks up the speaker – it currently supports eight languages covering about 90% of the world’s speakers – it sends the audio to an IBM Watson powered AI engine that completes the translation and sends it back. The round trip, said May, takes a three to five seconds over a decent cellular data connection, assuming the person is speaking clearly.
Current language support includes Mandarin Chinese, Japanese, French, Italian, German, Brazilian Portuguese, English and Spanish. Arabic is planned next with Cantonese also on the cards.
“We’re working on a few proof-of-concepts with some international companies. We’re looking to utilise the technology to do it for realtime phone conversations over VoIP,” said May.
The journey from May’s original idea through to going to market this week has taken four years.
“I’ve had a lot of advisors along the way. When I started this, it was initially an ear-piece. One of the advisors was a high-up engineer at Google. His advice was to move away from the hardware side and focus on the technology and get that in order and then move towards the ear-piece which is what we’ve done”.
This approach means Lingmo really has two products. The ear-piece, which they’ll sell for $219 (US$179) and access to their translation APIs and services. If you buy the headset, there are no ongoing fees for access to the translation system.
As the TranslateOne2One ear-piece has a SIM card, it can also be used to make calls. The idea is that when you land overseas, you buy a local SIM card, plug it into the TranslateOne2One and you’ll have one device that can make and receive calls, and translate when people speak to you.
In order to hold a conversation, both you and the other person require a TranslateOne2One. Lingmo has thought of that by selling a two-pack with one device designated as the master and the other as slave.
The 1.5-inch LCD screen, on the side, can be used to display directions from Google Maps. It’s also used to tell the device which languages it will be working with so that the AI, doing the translating, has one less thing to do.
May says that despite all of this, the TranslateOne2One weighs about 32g. It hooks around both ears and is designed so people can wear hats or glasses while it’s on.
The relationship with IBM came through an introduction between May and IBM Master Inventor Neil Sahota. That led to discussions that ultimately resulted in a partnership that utilised Watson’s natural language recognition capability.
Lingmo’s business model is interesting. While the TranslateOne2One ear-piece seems quite inexpensive, the data it collects allows Lingmo to further optimise its translations APIs which improves the services they sell through their APIs.
“It’s not just about making as much money as we can from the ear-pieces. it’s about the data collection. The more people that use it, the better it gets trained, the better the accuracy and the better the latency time,” said May.
The plan is to further refine the ear-piece over time and develop a smaller “series 2” device.