A survey on construction and enhancement methods in service chatbots design

Abstract

Chatbots are being widely applied in many service industries to help schedule meetings, online shopping, restaurant reservations, customer care and so on. The key to the success of the service chatbots design is to provide satisfying responses to the given user’s requests. This survey aims to provide a comprehensive review of chatbots construction and enhancement methods. We first introduce major techniques for the three core design philosophies, which are rule-based, retrieval-based and generation-based methods, followed by a brief summary of the evaluation metrics. Then we present methods to enhance service chatbot’s capabilities with either an ensemble of multiple chatbots, collaborating with human workers or learning from users. Finally, in future directions we discuss the promising response generation models for chatbots using the recent progress in the transformer and contextual embeddings, as well as potential ways to construct a chatbot with personality to achieve a better user experience.