Date Published: August 08, 2012 - Last Updated 5 Years, 108 Days, 2 Hours, 57 Minutes ago
The use of speech analytics has become commonplace for quality management because it lets the client find, filter and interpret relevant calls through a goal-oriented, automatic solution, unmatched by frequently aimless human searching. For contact centers, calls from unsatisfied customers, found through emotion detection, seem to represent an especially convenient and promising approach with a high potential for quality improvement. But, further analysis is needed.
The Disparity and Complexity of Emotions
By definition, the labeling of emotions represents a highly subjective process, especially when detected in an automated manner. Emotions are conveyed in many different ways such as voice, gesture, facial expression and body language, and they are often affected by past experiences as well.
Emotions will never be detected with 100 percent accuracy using scientific methods due to the human factor involved. Furthermore, different people express their emotions in different ways: some may shout while others stay calm and express their dissatisfaction factually. In addition, some people speak slowly and softly by nature while others always speak loud and fast. And people of different geographical regions tend to use different styles of speaking.
These cultural and human elements suggest barriers to the use of emotion detection to improve contact center quality, and they make it look like a gimmick, designed primarily to enrich the developers.
Emotion Detection for Contact Centers
To analyze further, we must first define the goals of an emotion detection solution in contact center environments. Should it detect many different emotions without fail? Some systems today can automatically spot emotions such as anger, happiness, sadness and disgust or a lack of any strong emotions at all. But contact centers primarily need to filter out calls with unsatisfied or angry customers. Only by finding and analyzing these problematic interactions can they determine the reasons for customers’ unhappiness and take steps to fix them. And if these calls are detected in time, the contact center may still be able to avoid the loss of the original angry customer.
Thus, contact centers really do need a solution to automatically detect anger calls out of hundreds or thousands of daily interactions. But they do not need to detect other emotions. Thus, analysis is facilitated for the emotion detection engine because the more emotions a system must detect, the greater the probability of error.
In addition to the variability of feelings, some solutions must find very subtle emotions. But this requirement represents another advantage for contact centers because anger provides unique signals and has been proven by scientific studies to be automatically detected better than any other emotion.
So, the special emotion detection demands of contact centers lend themselves to an automated solution.
Improving Emotion Detection
Of course, an ideal speech analytics solution does not exist. But contact centers do not need to find every single anger call when an adequate sample can be analyzed in depth, and appropriate general measures can be taken to improve quality.
However, for those requiring a higher level of accuracy, emotion detection can be combined with keyword spotting, another capability of speech analytics. For example, sometimes curse words indicate anger as much as raised voices.
Keyword spotting provides many other benefits as well, and, in addition to emotion detection, it gives one more indication of the value of speech analytics for today’s contact center.