Courteously Yours: Inducing courteous behavior in Customer Care responses using Reinforced Pointer Generator Network

Proceedings of NAACL-HLT 2019, pages 851–860 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 851 Courteously Yours: Inducing courteous behavior in Customer Care responses using Reinforced Pointer Generator Network Hitesh Golchha∗, Mauajama Firdaus∗, Asif Ekbal, Pushpak Bhattacharyya Department of Computer Science and Engineering Indian Institute of Technology Patna Patna, India (hitesh,mauajama.pcs16,asif,pb)@iitp.ac.in Abstract In this paper, we propose an effective deep learning framework for inducing courteous be- havior in customer care responses. The inter- action between a customer and the customer care representative contributes substantially to the overall customer experience. Thus, it is imperative for customer care agents and chat- bots engaging with humans to be personal, cordial and emphatic to ensure customer sat- isfaction and retention. Our system aims at automatically transforming neutral customer care responses into courteous replies. Along with stylistic transfer (of courtesy), our sys- tem ensures that responses are coherent with the conversation history, and generates courte- ous expressions consistent with the emotional state of the customer. Our technique is based on a reinforced pointer-generator model for the sequence to sequence task. The model is also conditioned on a hierarchically en- coded and emotionally aware conversational context. We use real interactions on Twitter between customer care professionals and ag- grieved customers to create a large conversa- tional dataset having both forms of agent re- sponses: generic and courteous. We perform quantitative and qualitative analyses on es- tablished and task-specific metrics, both au- tomatic and human evaluation based. Our evaluation shows that the proposed models can generate emotionally-appropriate courte- ous expressions while preserving the content. Experimental results also prove that our pro- posed approach performs better than the base- line models. 1 Introduction With the advancement of artificial intelligence (AI) and natural language processing (NLP), au- tomatic systems have made immense impact on human lives by assisting them in their everyday ∗ First two authors are jointly the first authors works. Human-computer interaction is pervasive in many applications such as chatbots, personal assistants and many more. Natural language gen- eration (NLG) component of such systems is an important aspect of every human computer inter- action. Thus research in recent years have been on modulating biases, styles and control in text gen- eration to enhance these interactions. Customer care is an essential tool used by companies to provide guidance, assistance and in building stable customer relations. The ease of ac- cess, ease in following-up and immediacy of so- cial media has made it a strong platform for com- panies and applications to interact with their cus- tomers. In this platform, we see the usage of cour- teous and emphatic language, which is the cen- ter of our current study. For the growth of any company or application it is necessary for the cus- tomer care agents to be cordial and amicable to the customer. Thus along with handling queries, it is important for agents to provide customer satisfac- tion by greeting, empathizing, appreciating feed- back, apologizing at the right time, and thus build a …

Content truncated to 500 words