This version - version 2.
Devised by Christopher Newell
Written by Christopher Newell and Avani Shah
Original version - version 1.
Developed with Paul Elsam, Tim Howle and Ben Johnson
Background: Actors and musicians are silent when fitting? Machines are silent when broken?
Any speech whether human or computer generated will be punctuated by pauses. It will also be perceived in a place that may be fitting or absurd. The place will penetrate the silences negating them and adding noise. Silences suggest engagement. Silences hint at interaction when none exists.
Any voice will tell a story: a story that articulates the words and the speaker.
For a synthetic actor the words, the speaker and the place matter no less than for a human actor.
No voice can be neutral. No computer voice need be natural. Appropriateness may be measured by engagement and engagement in this experiment means staying on the line for 3 and a half minutes.
Research: According to Goldman Eisler (Eisler, 1968) between 4% and 54% of speech in an interview situation is silence. The silences may be aligned to grammatical requirements or occur unpredictably. We exploit this fact to synthesise spontaneous pausing and hesitations in computer speech in an effort to improve listener acceptance. Will the listener listen for longer and more carefully if we embed random duration pauses (between sensible limits) at full-stops and commas?
Eisler, F. G. (1968), Psycholinguistics: experiments in spontaneous speech, London, Academic P. 1968.