Characterising dialogue call-flows for pervasive environments

Abstract

Algorithm design and analysis rapidly attained maturity with the introduction of time and space complexity measures. Dialogue call-flow design and analysis for pervasive devices is a relatively nascent art, complicated by the variation in resource constraints. We investigate the problem of dialogue call-flow characterisation for pervasive devices, with the objective of defining complexity measures for dialogue call-flows. A dialogue call-flow executing on a pervasive device is an interplay among the device, the human and the speech system, and can therefore be completely characterised by the 〈resource,usability,technology〉 triple. Topical examples of resource are memory and energy; usability is indicated in the number of questions; and technology is exemplified in the accuracy of the speech recognition system. These are charateristics of the call-flow, but the call-flow characterisation itself is {device,human,speech system}-independent. We instantiate 〈r,u,t〉 with 〈memory,questions,accuracy〉 to introduce the 〈m,q,a〉-characterisation of dialogue call-flows. We select m, q, a metrics to define various 〈m,q,a〉-complexity measures for a call-flow. Every call-flow thus has a complexity measure associated with it - a feature indispensable for the systematic analysis and design of dialogue call-flows.