A Two-stage Attention-based Model for Customer Satisfaction Prediction in E-commerce Customer Service

Anonymous

Abstract: Nowadays, customer satisfaction prediction (CSP) on e-commerce platforms has become a hot research topic for both intelligent customer service and artificial customer service. CSP aims to discover customer satisfaction according to the dialogue content of customer and customer service, for the purpose of improving service quality and customer experience. In this paper, we focus on CSP for intelligent customer service chatbots. Although previous works have made some progress in many aspects, they mostly ignore the huge differences of expressions between customer and customer service, and fail to adequately consider the internal relations of those two kinds of personalized expressions. Thus, for emphasizing the importance of modeling customer part and service part separately, in this work we propose a two-stage dialogue-level classification model, which contains an intra-stage and an inter-stage to handle the issues above. In the intra-stage, we model customer part and service part separately by using attention mechanism combined with personalized context to obtain {\it customer state} and {\it service state}. Then we interact those two states with each other in the inter-stage to capture the final satisfaction representation of the whole dialogue. Experiment results demonstrate that our model achieves better performance than several competitive baselines on our in-house dataset and four public datasets.

Paper Type: long