In recent years, machine learning (ML) has been an increasingly significant role in customer dissatisfaction analysis using customer comments (CCs). Although promising, existing ML methods suffer from ineffectiveness due to the incompleteness of information obtained solely from CCs. Another type of information pertaining to professional service aspects— agent notes (ANs), introduced by the agents on behalf of enterprises—effectively complements the weakness of CCs. CCs and ANs represent evaluations of customers and agents on the same service from different perspectives, providing a comprehensive view of the service quality. However, it is challenging to directly apply ML methods to integrate CCs and ANs with complex bilateral relationships to generate explainable and actionable insights for customer dissatisfaction management.
This paper examines dissatisfaction resulting from the gap between customer expectations and delivered service quality and proposes a novel, explainable method based on domain knowledge for analyzing and tracing customer dissatisfaction through the service quality gap from bilateral CCs and ANs in a comprehensive way. Furthermore, to trace dissatisfaction in an explainable manner, an information contribution indicator measure () and inference strategies are devised to help map the potential gap dimension to the service quality dimension metrics and trace the informational links in the network. Extensive experiments reveal that the proposed XGAP method is advantageous over baseline methods in learning performance as well as explainable tracing.
扫码关注我们
求助内容:
应助结果提醒方式:
