Sixth European Conference on Speech Communication and Technology
(EUROSPEECH'99)

Budapest, Hungary
September 5-9, 1999

OASIS - a Framework for Spoken Language Call Steering

Mike Edgington (1), David Attwater (2), Peter Durston (2)

(1) now at: SRI International, Cambridge, UK
(2) BT Laboratories, Ipswich, UK

OASIS is a research project at BT Labs investigating practical large-scale spoken language automation of call steering (call routing). BTís own operator assistance service is used as an initial trial domain. Spoken language call steering requires understanding of both userís language and behaviour. Therefore, the OASIS project makes extensive use of corpus-based machine learning techniques and iterative Wizard-of-Oz simulations to verify and refine the system design. We also describe a simple four layer model of general telephone call-handling, developed to characterise services in order to identify appropriate automation technology. We argue that for many helpdesk applications, a high proportion of callers will just describe the nature of their problem, rather than refer to particular services offered. For these calls, language use is highly unpredictable, and corpus based learning techniques are essential for speech recognition and understanding.


Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Edgington, Mike / Attwater, David / Durston, Peter (1999): "OASIS - a framework for spoken language call steering", In EUROSPEECH'99, 923-926.