The Business Case for Speech Translation in UC
In my previous two blogs I've written about Skype Translator and the general technology framework behind the service. Now let's talk about how all this might apply to unified communications.
Most everyone immediately considers the value of having UC-enabled speech translation available in the call center. It seems natural, since call centers have been evolving to leverage UC and the need to provide multi-language assistance to callers is expensive and fairly complicated to manage. The problem here is that 95% accuracy level I described in last month's blog.
Ninety-five percent accuracy is not good enough for conducting contract negotiations or other communications where precision is a business necessity. Is 95% good enough for a call center? The answer depends on the level of precision and/or accuracy needed to support the transaction, and that is not just a function of technical accuracy but also how the level of "service" might be impacted by the 5% error rate in mistranslation.
The field of linguistic gestalt has already shown that most people have the cognitive capability to fill in missing pieces of a conversation. Therefore, it is likely that we will grow to tolerate imprecision that does not materially impact the transaction. It is also likely that, while language accuracy will generally struggle to exceed 95%, the predictable "language" of many kinds of service calls will result in better specific accuracy around order-taking, tech support, and other transactional activities. Legal and social criteria may not allow UC-enabled speech translation for 911 calls or premium-tier services, but the flexibility of UC-translation may be acceptable even for these in back-up situations.
So my predication is that, yes, services like Skype Translator are going to find their way into call centers, and probably within the next couple of years. The same is true for front-desk, help-line, medical assist, and similar services. I was recently in Taiwan and would have loved to have been able to order room service without having to walk down to the lobby and point to menu items because they couldn't understand my English and I could not understand their Mandarin.
A second major application will be in collaboration, particularly conference calls. For example, my Spanish speaking is little better than 60% accurate and my listening well below 50% and, while, my colleagues in Peru are better with English, they are only marginally so. At 95% accuracy, a product like Skype Translator would substantially enhance the quality of our conference calls. Not only will UC-enabled speech translation find its way into Microsoft, Google, and Cisco enterprise UC products, but it is very likely that conferencing services like PGi, InterCall, and those offered by the Telcos will very shortly offer speech translation services.
Related products in education and training will likely embed UC-enabled speech translation as a way to add value to products that can be used to cross language and location barriers between teachers and students. And the inevitable UC services from Facebook and LinkedIn will benefit from speech translation.
Where I see UC-enabled speech translation really hitting it out of the park is around how it will empower individual consumers to overcome language barriers to explore the world. Private pilots could fly into countries and talk to control towers where before language was a barrier (if 95% is good enough for aviation). Vacationers could call remote foreign hotels directly and talk with proprietors with confidence. Second generation children of immigrants who never learned the mother tongue could call their grandparents.
To be sure, speech translation has been around for a while (e.g., SpeechGear, Jibbigo, Blabber) but they lacked the platform that UC provides, and these earlier products did not have the product muscle of Microsoft, Google, and Cisco. At 95% speech translation accuracy, the market is ready to go.
I hope that the UC vendors don't mess up this opportunity - the speech translation user experience needs to be simple and easy. Wrapping it in complicated or expensive licensing is just as bad as blowing the UX. If the UC vendors do mess this up, in a couple of years someone is going to release a mobile-phone app or wearable speech device that won't rely on or leverage the UC platform at all. The moment for UC vendors to get this right is right now.
Also on UCStrategies.com on this topic: