When people talk to each other, they can get a better idea of the moods coloring their speech by looking at their inflection, tone of voice, body language and the like. But machines like IVR—Interactive Voice Response—platforms enjoy no such capability. Though that's looking to change with some new advancements in the field.
Perhaps the biggest problem in bringing tone and inflection capabilities to IVR systems is that there are very few standards from which to operate. If humans worked like, for instance, the Elcor from the popular Mass Effect series of games, in which they announced their mood at the beginning of every sentence, it would be much easier to build those kinds of capabilities into IVR systems.
But with humans, it's a much different field. For instance, some people who talk quietly do so because they're feeling sad or tired. Others who speak quietly do so because their rage is mounting and they're trying desperately to control themselves from breaking down into a quivering mass of fury. For an IVR to assign one parameter to a sudden use of lowered volume would be next to impossible, and would likely prove disastrous; imagine if the IVR interpreted the volume drop as “tired” for the person for whom it signifies “mounting rage”.
Thus, most standard types of machine language, which depend on the “if-then” relationship will find themselves at least somewhat stymied by the human use of tone and inflection. However, so-called “speech emotion recognition systems” are working to bridge the gap between human emotion and its sheer lack of standards. In such a system, the machine receives speech input, and then the recognition software dissects the exchange, analyzing and cataloging it for the best shot at interpreting just what was meant, despite what was said.
Emotional states can then be at least somewhat classified according to different variables within speech that can be cataloged and classified, including things like energy, pitch, and duration. This in turn can allow the IVR to respond differently according to the speech being placed to it.
As for the accuracy of such a system, that's still being tested. But an IVR that can respond to human emotion would be a valuable marketing tool indeed. Imagine a system that can field angry calls, determine the level at which the caller is angry, and then work to defuse that anger with special offers. Many of us have had our own experiences in which we began angry at a company for failing to provide some service that was promised, or failing to do a proper job of providing same, but then had that anger replaced with satisfaction when discounts or credits were brought into play.
It bears watching, of course, to see just how well machines can both recognize and respond to human emotion. But with such capabilities in hand, any firm using an IVR stands to benefit substantially.
Edited by Juliana Kenny