Development of Chatbot Retrieving Fact-Based Information Using Knowledge Graph

Abstract

Chatbot assists users by providing useful responses and not just a conversational system functionalities. The advanced Chatbots such as Siri and Alexa are the results of evolution of different response generation and NLU techniques that have arrived since 1960s. Usually, chatbots are designed to address domain-specific queries; for instance, a medical chatbot requires the user to provide his/her symptoms; in the corporate world, the chatbots designed are mainly for addressing the FAQs asked by their clients/customers. However, state-of-the-art technologies are emerging, and knowledge graph is one of them. The idea of using knowledge graphs is that the data stored in them is linked. The proposed-chatbot addresses the problem of answering factoid questions by retrieving information from knowledge graphs. Initially, the neural machine translation approach was used; however, due to its limitations, keyword extraction approach was adopted for the proposed chatbot. In order to compare the proposed chatbot system with DBpedia metrics, the F-measure quality parameter were used for determining the overall performance of chatbots.