Development of Julius-based speech dialogue system for campus receptionist robot

摘要

ASKA is a speech-orinted humanoid robot system to realize an automated receptionist agent of a university. The speech dialogue function is composed of speech understanding module and text-to-speech synthesis module, each communicates to other parts such as object sensing module and genture generation modules to accomplish an distributed system integration. The recognition scheme is based on LVCSR engine Julius, with task-dependent N-gram language model and keyword-based meaning extraction algorithm. The development is a collaborative work of several laboratories in NAIST, and is aimed to be a common verification platform of various agent technologies and engineerings in real environment. In this paper, we describe the overview of ASKA, the implementation method of the speech dialog functions on ASKA and the future schedules.