A call center is a primary channel for service encounters in a variety of industries, including airlines, hotels, and financial and telecom services. According to a recent industry report, calls to businesses were expected to exceed $169 billion per year by 2020. The quality of customer services provided by call centers is crucial in determining customer satisfaction. A poor service quality can lead to frustrated customers, bad word-of-mouth, and customer switching. To ensure service quality, call centers usually use two approaches to manage service quality: sampling calls for manual inspection and customer call backs. However, both approaches are costly and labor intensive. As a result, companies can only afford to check a small number of calls for potential quality problems, often long after the service encounters are over. Needless to say, cost-effective and scalable approaches for monitoring service quality are of great value to companies.
This project tests an automated approach of managing service quality based on real-time speech emotion recognition. First, the researchers extract acoustic and textual data from customer calls, then apply several machine-learning/deep-learning technologies to detect negative customer emotions, and finally use the recognized emotion sequences to detect service quality problems and provide real-time intelligence to customer service representatives and QA specialists in call centers. With this automated, machine-learning based approach, the researchers hope to bring forth a scalable, cost-effective solution for call service quality management.