SpEx: A Tool for Visualising and Navigating Speech Audio

Abstract

Audio is a ubiquitous form of information that is usually treated as a single, unbreakable, piece of content. Thus, audio interfaces remain simple, usually consisting of play, pause, forward, and rewind controls. Spoken audio can contain useful information across multiple topics and finding the information desired is usually time consuming. Most audio players simply do not reveal the content of the audio. By using the speech transcript and acoustic qualities of the audio, I have developed a tool, SpEx, which enabled search and navigation within spoken audio. SpEx displayed audio as discrete segments and revealed the topic content of each segment using mature Information Visualisation techniques. Audio segments were produced based on the acoustic and sentence properties of speech to identify topically and aurally distinct regions. A user study found that SpEx allowed users to find information in spoken audio quickly and reliably. By making spoken audio more accessible, people can gain access to a wider range of information.