Simulation and Recognition of Information Dialogues using Markov Models

title: Simulation and Recognition of Information Dialogues using Markov Models
author(s): R.J. van Vark, J.P.M. de Vreught, L.J.M. Rothkrantz
published in: 1996
appeared in: Proceedings of the 8th European Simulation Symposium and Exhibition
Genova, Italy
pages: pp. 533-538
download: PostScript

Abstract

OVR, a Dutch provider of information concering national public transport, has started an investigation of the usability of Automated Speech Processing to automate part of their services in order to be able to serve all customers satisfactory. Meanwhile, a corpus recorded during this research is used to develop a training simulator for training their human operators to handle conversations more efficient. Based on the corpus a coding scheme was developed to analyse public transportation information dialogues. To gain better insight in how dialogues are taking place, Markov models are developed using this coding scheme. These models are used to cluster dialogues and to develop scripts which ideal dialogues must follow. The Markov models derived to be used in the training simulator can later be used in the dialogue management component of the ASP system.

 
blue line
University logo