AFDM : an Analytical Framework of Dialogue Manager

Abstract

A computer system which provides the possibility of human- like conversation with the system for users, is called dialogue system. A dialogue system consists of different components. Dialogue manager is the core component of every dialogue system composed of Dialog State Tracker and Policy Learning. A dialogue management system is able to direct a conversation between agents, including both human and computer. A variety of models have been developed for dialogue manager. This paper provides an Analytical Framework for Dialogue Managers (AFDM) which overview wide range of techniques applied for developing dialogue manager component. AFDM framework works in three parts including: 1) a classification of current approaches applied to dialogue management task, 2) introducing the evaluation measures and 3) analytical evaluations. Also, we discuss the main qualitative properties of each approach. This framework is designed in order to pave the way for future research with the aim of a) improving current models, b) selecting a method based on its properties, or c) comparing future strategies.