A next generation chatbot-framework for the public administration

Abstract

With the growing importance of dialog system and personal assistance systems (e.g. Google Now or Amazon Alexa) chatbots arrive more and more in the focus of interest. Current chatbots are typically tailored for specific scenarios and rather simple questions and commands. These systems cannot readily handle application domains characterized by a large number of relevant facts and complex services (e.g., offered by the public administration). In this work, we present a chatbot framework tailored specifically to the needs of public administrations able to provide answers for all types of questions related to offered services and offices. The challenges in this scenario are the large number of relevant services, the complexity of administrative services, the context-dependent relevance of user questions, the differences in expert-language and user-language as well as the necessity of providing highly reliable answers for all questions. We present the developed framework and discuss our experiences obtained while running the chatbot publicly with real users. The results show that our system efficiently provides relevant answers to the user questions. We explain how our system extends existing approaches and discuss the experiences with the live system.