IntelliBot: A Dialogue-based chatbot for the insurance industry

摘要

Chatbots have become the go-to platform for users to receive answers to their queries. They are now being used by many businesses too to provide their customers with a virtual assistant to answer their queries. But when it comes to engaging with a user in a dialogue, existing chatbots have several shortcomings, with issues such as failing to provide a meaningful response to the user, offering semantically incorrect information etc. This paper studies the working styles of existing chatbots in generating a response and then identifies their shortcomings from the viewpoint of engaging in a dialogue with a user. It then proposes a domain-specific chatbot named IntelliBot, which is a response-generating dialogue-based chatbot that uses multiple strategies to generate a response. IntelliBot was trained on two datasets, namely the Cornell movie dialogue and a custom-built insurance dataset so it has domain-specific knowledge. The performance of IntelliBot was then validated and compared with three other chatbots from the literature, namely RootyAI, ChatterBot and DeepQA. The results demonstrate IntelliBot's superiority in engaging with the user and providing a complete answer in the insurance domain. (C) 2020 Elsevier B.V. All rights reserved.