A Multi-Agent Reinforcement Learning Algorithm for Disambiguation in a Spoken Dialogue System

A spoken dialogue system (SDS) communicates with its user(s) in a spoken natural language. It responds to user speech input for answering questions, providing advice, and so on. Correctly understanding user input is very important to system performance. A key issue in understanding user input is handling ambiguity since any natural language is ambiguous. In our research, we develop a novel multi-agent reinforcement learning algorithm for disambiguation in a spoken dialogue system. In the algorithm, multiple agents learn knowledge about user behavior in activities and language use, and the knowledge is used to handle ambiguity. In this paper, we introduce the multi-agent reinforcement learning algorithm, and describe a spoken dialogue system for mathematics tutoring that we build to implement and experiment the algorithm.