ISCA Archive Eurospeech 1993
ISCA Archive Eurospeech 1993

Methods for traversing a pre-recorded speech message network to optimise dialogue in telephone answering systems

Mary Zajicek, Ken Brownsey

This paper addresses problems associated with conversations with computerised telephone answering systems. Using these systems, we often find ourselves listening to lengthy menus, forgetting the number of our desired option, choosing the wrong one and being unable to get back without hanging up completely, and wasting valuable time and money listening to inappropriate messages. In addition, user acceptance of this form of interaction is a significant problem[l]t and user's correct anticipation of what is required of them is difficult to manipulate [2], [3] This paper describes an approach to a dialogue enhancing techniques prototyped by the authors. Our aim was a) to make dialogues more flexible b) reduce the number of messages that needed to be listened to and c) reduce the number of unsatisfactory outcomes, i.e. when users do not achieve the objective of the phone call Most telephone answering systems comprise of a hierarchical menu system made up of pre-recorded speech messages which are accessed by the caller using a small speaker independent, vocabulary of words[4]. The dialogue structure can be viewed as a network of messages which must be traversed in order to reach action nodes which will enable the caller to either a) retrieve an item of pre-recorded information b) leave a message on an answering machine, or c) talk to the relevant human operator. In a standard hierarchical menu system - see FIG(1), network traversal is one directional, and between adjacent nodes. Our dialogue enhancing techniques enable the user to "jump" from one message node to another more appropriate one in a completely different part of the network. The techniques discussed are: - Simple backtracking - Heuristic relocation via keywords. - Direct access via keyword - Ternary choice

Keywords: Cognitive Model, Dialogue, Heuristic, Network Traversal, Non-hierarchical, Speech, Usability, User Model