Tuesday, August 9, 2011

Talk To Me Baby

Over ten years ago I was fascinated by the introduction of a new service called Wildfire. It was a totally automated telephone assistant that understood natural language speech. Your virtual assistant could screen your calls, find you and interact with you, allowing you to decide where and when you wanted to speak to the person calling you. In the demonstration Wildfire would also handle your calendar, reminding you of appointments, moving them or making new ones. You could command Wildfire to place calls for you, too.

No software or hardware was required. Today, we would call it cloud-based. Wildfire was configured, controlled and operated with voice commands. Unlike most other voice activated systems of the day, it required no training. You did not have to teach it how to understand your voice. By the way, it never raised its voice or called in sick on Monday morning.

Voice technology has steadily improved over the years. We are all too familiar with customer service organizations that use interactive voice recognition (IVR) systems. Press 1 for English. Say or speak your 14 digit account number. Your card has been activated; if you need further assistance please say "YES." Computers routinely talk to us by phone and guide us through inquiries, transactions and other activities. While reviewing this column this morning, as if on queue, I received a phone call from the local car dealer confirming my service appointment. "Hello," the pleasant automated voice said, "We are confirming your appointment for twelve o'clock ... if you are still planning on coming, please say yes."

While we haven't quite reached the level of quality of the Enterprise from Star Trek or HAL from 2001 Space Odyssey, the computer has gained the ability to understand and respond in a wide variety of situations. Voice recognition is included in the Windows operating system. Adobe Reader will, in fact, read a document to you. Google Voice will transcribe your voice mail messages and email them to you . Smart phones (iPhone, Android) will execute a spoken search request and offer other applications that readily understand verbal commands. Hands free capabilities in our cars initially allowed us to place phone calls by voice command. Now we control the entertainment system, navigate, book a restaurant and adjust the climate controls as well.

Voice activated systems are all around us. So why hasn't voice activation achieved broader acceptance? One of the last remaining hurdles is the need for a trigger. To get its attention you had to say, "Wildfire" before speaking a command. In your car you typically push a button and, after the tone, speak your command. Voice activation must advance to the point where your commands are heard in context and allow the system to realize when you are speaking to it.

This is no small feat. Imagine if your television was always listening. What should it do if it hears you say, "let's see what's on channel seven?" Does this mean tune to that channel or display the guide? How would it distinguish this from a comment directed at another person in the room. For example, "I don't like this show you are watching, let's see what is on channel seven." Before executing any command, the computer may have to engage in conversation. At a minimum it should confirm your request to eliminate ambiguity. "Did you want me to put on channel seven or show you the guide?" Software usually protects us from ourselves by confirming a file delete or asking if we really want to quit and lose our changes. This would be even more critical in situations where spoken words could have serious unintended consequences.

There is some work being done in this area but until it becomes mainstream I'm afraid we're going to have to continue to somehow get the computer's attention before asking it to do something.

Most wives will understand this since husbands typically operate the same way.

Captain Joe

Follow me on Twitter @JPuglisiLLC