One of the most fundamental changes to how you will be able to do business in 5 years is nearly here. Instant voice language translation will enable you to have a conversation with someone with both of you speaking your native language. This opens up who you will be able to do business with, breaking the barrier of language communication.

For anyone who has read ‘The Hitchhiker’s Guide to the Galaxy’, they will know about the Babelfish, a tiny animal you put into your ear, letting you understand someone speaking any language in the universe, and letting them understand you when you answer in your own language.

And thanks to a breakthrough from Microsoft’s R&D Lab being tested in Skype, this ability will soon be accessible to you. Beta testing is likely to happen later this year.

The breakthrough is that the system is now fluent and reliable enough for reasonable usage. Enough being the key word here. When people learn a foreign language, there comes a point quite early on when you have a minimum vocabulary and grammar to go from not being able to interact with someone in another language to being able to share information, requests and opinions. In my own case, two years ago I went travelling through South America, and although I learned probably less than 200 words in my lessons, that was enough for me to be able to navigate independently in unfamiliar places.

As you can see in the video above, the Skype system is already significantly more capable than that. Using it, people can go from not being able to collaborate across a language barrier to having the sort of basic conversations which lead to productive outcomes. And as the system gets updated automatically in the cloud, accuracy rates will continue to climb.

I see this having a big impact in the next 5-10 years in three main domains:

  • Collaboration: imagine individual team members with specific skills and knowledge being able to work on a project no matter where they were based. Within international companies, this can often mean people in various countries with various native languages. Virtual teams could therefore be put together based on individuals skills, instead of only selecting people who would be able to speak the same language. Worldwide conference calls could benefit even more.
  • Knowledge sharing / teaching: there is so much knowledge out there which is unobtainable for various people because they cannot understand it. Distance learning could benefit greatly from such a system. Imagine if an expert in bio-molecular chemistry could give a public lecture at a university in São Paulo and have it simulcast to universities in Beijing, Montreal and Prague, with students able to ask questions instantly. Even large companies could broadcast their CEO’s annual speech to every office simultaneously.
  • Doing business: what I am most excited about! Imagine that you are looking for a new supplier or manufacturer who could be anywhere in the world. You get some recommendations of a few companies in China, India, Russia and Egypt. Using such a system (assuming it has spread to their countries), you would be able to call each of these companies and speak to someone directly to find out what their capabilities and prices were to help inform your decision. Ultimately, the technology could even be used in a face-to-face meeting using a phone connected to the cloud, translating each sentence back and forth.

One caveat I do have though is that I think that the system would only work via a medium like Skype, where reading facial expressions give each person instant feedback on whether what was translated made sense or if the other person seeks clarity. After all, at the moment the voice output is still a computerised voice, and in practice this would lead to higher levels of frustration.

Ultimately, just as the telephone and internet broke down the barriers to sharing information across borders, I believe this technology will enable the next generation to break down the barriers of communication caused by language.

The previous problem with translation

Classic case of poor translation

Classic case of poor translation

The problem which has plagued both speech recognition and translation by software in the past is that computers need to match inputs to the most likely ‘correct’ result, based on comparing the input to everything it has in its database. Software is extremely good at finding known, exact matches, such as having confirmed that the words “I love you” in English match with “Ich liebe dich” in German.

But computers find it much harder to deal with input that has even a bit of ambiguity. For example, while a human can instantly tell the difference between ‘underwear’ and “under where?” by slight nuances in their sounds, computers would find it extremely difficult to understand the difference of this input, and therefore selecting which input then needs to be compared to everything in a database to compare. The way a computer might do this is by taking a digital copy of the book “War and Peace” in several languages and comparing the sequence of words and sentences to find matches between languages. It is an unreliable and error-prone method, requiring a huge amount of computational power.

Microsoft’s translation breakthrough

But the team in Microsoft’s Research and Development laboratories (working on a huge number of exciting projects) have made a breakthrough which could soon make instant translation a possibility. In the latter part of 2009, they decided to invest in a new field of research which attacked the challenge using something called Deep Neural Networks. Don’t ask me exactly how it works, but it helps computers process information in a manner more similar to how humans do, and the net result has been a huge improvement in both voice recognition and translation.

Bring the two together, and add a computerised voice reading the translated text, and it becomes possible to translate what you say into multiple languages, and then back again. Using this system, in 2011 error rates dropped from about 1 in 4 words to about 1 in 8. It’s not 100% accurate yet, but as I said earlier, it’s good enough to get across the main point people were trying to make, and that is enough.

Have you had any funny experiences with translation errors in business? And when do you think that you would trust a computer to translate your conversation? Let me know in the comments below.

improvidesbookimprovementcovershadowGet a Free copy of my new eBook: The Secrets of ongoing Innovation Success and innovation insight every week by signing up for my mailing list.