TR-IT-0012 :1993.8.30

Christian Boitet

Integration of Heterogeneous Components for Speech Translation: the "Whiteboard" Architecture and an Architectural Prototype

Abstract:A new software architecture for Speech Translation systems, based on the use of a "whiteboard", is presented. Unlike a blackboard, the "whiteboard" is accessed only by a "coordinator", and not by the "components'', such as speech recognizer or syntatic analyzer. The main advantage of this architecture is to allow easy integration of existing or new components, without having to modify them in any way. Graphic examination of the state of the entire system can also be provided by the coordinator. An "architectural" prototype is under construction ans should be presented at the ATR workshop and the Waseda Symposium on Spoken Dialogue next November.

This report is an extension of a recent communication co-authored with M. Seligman [1]. Extensions concern essentially the rationalization of the kind of data structure proposed for the whiteboard, for which diagrams and examples from [2] have been used.

[1] Seligman, M. & Boitet, Ch. (1993) A "Whiteboard" Architecture for Automatic Speech Translation. Proc. International Symposium on Spoken Dialogue, Waseda University, Tokyo, 10-12 November 1993, to appear.

[2] Boitet. Ch. (1988): Representation and Computation of Units of Translation for Machine Interpretation of Spoken Texts. TR-I-0035 ATR, Osaka, 41 p.

The research reported here was conducted while I was staying at ATR Interpreting Telephony Research Laboratories, and then at ATR Interpreting Telecommunications Research Laboratories, as visiting researcher from GETA, IMAG, UJF&CNRS, France. I would like to heartly thank ATR for its constant support and very favorable research environment; and its members, from President through supervisors through researchers through secretaries to security personal, for all the personal help which they extended to me in so many ways at so many occasions.