Abstract Vanessa Pope 30 April 2018
While professional speech is often used as data for speech research, there has been little research into the craft of delivering speech to live audiences. Understanding how performers balance prepared material with flexible delivery benefits creative text-to-speech applications that aim for a conversational tone.
We propose and implement a subsequence matching methodology to identify how a professional stand-up comedian varies and repeats material across performances of the same show. The generalisable methodology highlights how a performer transitions into and out of prepared material by locating matches between performance transcripts, and between the transcripts and source text. Filled-pauses, false starts and repeated words are included in the transcripts to examine how apparent disfluencies interact with prepared material.
Initial analysis of five performances by one comedian found that transcripts of live performances are more similar one to another than they are to the source text, and that paralinguistic sounds are purposefully included in some repeated subsequences. Mapping consistent subsequences reveals aspects of show structure and identifies how subsequences are deliberately and spontaneously broken. The results show that filled-pauses occur disproportionately frequently immediately before and after matching subsequences, suggesting that disfluent speech could play a structural role in stand-up comedy as well as be part of the prepared material itself.
Vanessa Pope is a theatre director and PhD student in Media and Arts Technology at Queen Mary University of London, supervised by Prof. Elaine Chew and Dr Rebecca Stewart.