International Symposium on Chinese Spoken Language Processing (ISCSLP 2002)

Taipei, Taiwan
August 23-24, 2002

Intelligent Speech for Information Systems (ISIS): A Multi-modal, Trilingual, Distributed Conversational System with Combined Interaction and Delegation Dialogs

Helen Meng

Human-Computer Communications Laboratory, The Chinese University of Hong Kong

ISIS is a trilingual spoken dialog system in the stocks domain. It supports the three languages commonly used in Hong Kong (Cantonese, Putonghua and English), and serves as a test-bed for our research in various speech and language technologies. This talk presents the ISIS system with a focus on its several unique features. We use the CORBA middleware to implement a distributed architecture that supports interoperability across platforms. We incorporate KQML (Knowledge Query and Manipulation Language) software agents to handle delegation dialogs. We also implement a mixed-initiative dialog management strategy that combines online interaction with offline delegation. ISIS can automatically assimilate newly listed stock names into the systemís knowledge base. A recent enhancement supports multi-modal and mixed-modal input that suit the natural affordances of specific interactions in order to improve usability. Input modalities include speaking, typing or mouse-clicking. Output media include synthesized speech, text, tables and graphics.


Bibliographic reference.  Meng, Helen (2002): "Intelligent speech for information systems (ISIS): a multi-modal, trilingual, distributed conversational system with combined interaction and delegation dialogs", In ISCSLP 2002, paper INV3.