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

Liangqing Wu,Jiahao Ji,Zhoujun Li,Xiaodong He,Shuangyong Song,Zhenhe Wu,B. Zou

Abstract:Nowadays, customer satisfaction prediction 001 (CSP) on e-commerce platforms has become a 002 hot research topic for both intelligent customer 003 service and artificial customer service. CSP 004 aims to discover customer satisfaction accord-005 ing to the dialogue content of customer and 006 customer service, for the purpose of improv-007 ing service quality and customer experience. 008 In this paper, we focus on CSP for intelligent 009 customer service chatbots. Although previous 010 works have made some progress in many as-011 pects, they mostly ignore the huge differences 012 of expressions between customer and customer 013 service, and fail to adequately consider the in-014 ternal relations of those two kinds of person-015 alized expressions. Thus, for emphasizing the 016 importance of modeling customer part and ser-017 vice part separately, in this work we propose a 018 two-stage dialogue-level classification model, 019 which contains an intra-stage and an inter-stage 020 to handle the issues above. In the intra-stage, 021 we model customer part and service part sepa-022 rately by using attention mechanism combined 023 with personalized context to obtain customer 024 state and service state . Then we interact those 025 two states with each other in the inter-stage to 026 capture the final satisfaction representation of 027 the whole dialogue. Experiment results demon-028 strate that our model achieves better perfor-029 mance than several competitive baselines on 030 our in-house dataset and four public datasets. 031

Computer Science,Business