A Dialogue Model for Customer Support Services

Abstract

Many dialogue models have been proposed to learn the language model from the input queries for answering user requests. However, most models are not proposed for customer support services. Some shed light on answering user queries in a customer support system; however, they do not consider domain or emotion features implicitly hidden in user queries. In this study, we propose a deep learning framework to automatically answer user queries of customer support services. The proposed framework extracts domain and emotion features from user queries and then incorporates the extracted features into a generative adversarial networks model to generate the response to an input query. The extracted domain features may reveal user needs while the extracted emotion features may show the emotions implicitly hidden in the input queries. Therefore, the proposed model can better understand user requests and generate better responses. The experimental results show that our proposed framework outperforms the comparing methods and can generate better responses for user queries. Our framework may help companies provide 24/7/365 customer support services with less effort.