Abstract
Radio production is a creative pursuit that uses sound to inform, educate and entertain an audience. Radio producers use audio editing tools to visually select, re-arrange and assemble sound recordings into programmes. However, current tools represent audio using waveform visualizations that display limited information about the sound. Semantic audio analysis can be used to extract useful information from audio recordings, including when people are speaking and what they are saying. This thesis investigates how such information can be applied to create semantic audio tools that improve the radio production process. An initial ethnographic study of radio production at the BBC reveals that producers use textual representations and paper transcripts to interact with audio, and waveforms to edit programmes. Based on these findings, three methods for improving radio production are developed and evaluated, which form the primary contribution of this thesis. Audio visualizations can be enhanced by mapping semantic audio features to colour, but this approach had not been formally tested. We show that with an enhanced audio waveform, a typical radio production task can be completed faster, with less effort and with greater accuracy than a normal waveform. Speech recordings can be represented and edited using transcripts, but this approach had not been formally evaluated for radio production. By developing and testing a semantic speech editor, we show that automatically-generated transcripts can be used to semantically edit speech in a professional radio production context, and identify requirements for annotation, collaboration, portability and listening. Finally, we present a novel approach for editing audio on paper that combines semantic speech editing with a digital pen interface. Through a user study with radio producers, we compare the relative benefits of semantic speech editing using paper and screen interfaces. We find that paper is better for simple edits of familiar audio with accurate transcripts.