Automatic Satisfaction Analysis in Call Centers Considering Global Features of Emotion and Duration

Analysis of customers' satisfaction provides a guarantee to improve the service quality in call centers. In this paper, a novel satisfaction recognition framework is introduced to analyze the customers' satisfaction. In natural conversations, the interaction between a customer and its agent take place more than once. One of the difficulties insatisfaction analysis at call centers is that not all conversation turns exhibit customer satisfaction or dissatisfaction. To solve this problem, an intelligent system is proposed that utilizes acoustic features to recognize customers' emotion and utilizes the global features of emotion and duration to analyze the satisfaction. Experiments on real-call data show that the proposed system offers a significantly higher accuracy in analyzing the satisfaction than the baseline system. The average F value is improved to 0.701 from 0.664.