A Configurable Dialogue Platform for ASORO Robots

This paper is concerned with the architectural design and development of a spoken dialogue platform for robots. The platform adopts modular software architecture and event driven communication paradigm which makes speech enabled hardware devices and software components configurable and reusable. The platform is able to integrate heterogeneous dialogue components (such as speech recognizer, natural language understanding, speech synthesizer, etc.) through message-oriented middleware and a number of adaptors in a plug-and-play fashion. The dialogue system is empowered by a multi-purpose XML-based dialogue engine which is capable for pipeline information flow construction, programmable event mediation, multi-topic dialogue modeling and different types of knowledge representation. The proposed platform provides a generic framework for the easy and quick construction of robust, efficient and flexible spoken dialogue applications ranging from simple state-based dialogue prototype to complex frame-based and plan-based, mixed-initiative, multi-topic spoken dialogue applications. Most importantly, the system is cross platform, domain independent, service oriented, extensible through various plug-ins and capable of knowledge sharing within components. The platform has been deployed on various robots (robotic butler, smart home helper robot, receptionist, etc.) in A*STAR Social Robotics (ASORO) laboratory for the creation of different spoken dialogues. Results showed that components can be reused and configured for different robots without any programming effort while the development cycle was reduced significantly.