Time-compressed speech: Emerging ideas for audio in computer-based learning

Year: 1994

Author: Göllner, Winnie, Harvey, John

Type of paper: Abstract refereed

Recent advances in the capabilities of personal computers have now made high-resolution colour graphics, animations, video clips, and audio relatively commonplace in computer-based learning (CBL). While text clearly remains the major medium, programs are appearing which use speech either as a supplement to the text or, less frequently, as a replacement.

It is standard advice in interface design that filling the screen with text is undesirable-users simply do not read it. Yet the user must often be given such information. Designers thus tend to employ multiple screens, scrolling boxes, or condensed, point-form text. Systems which can offer speech as an alternative are therefore attractive as they not only free valuable screen space but a voice can carry nuances and cues lost in text. Further it is a relatively simple matter to provide alternative languages and voices.

The disadvantage of speech is demonstrated by Junor (1992) who estimates that listening to a message may take up to three times longer than reading it. Further, it is normally not possible to speed scan an audio track either to listen for the gist or to find a particular section.

Speech researchers have been interested in meeting these challenges by using time-compressed speech (TCS)-speech which by various means is speeded. Studies suggest that this has positive advantages. Arons (1993) reports both improved intelligibility and comprehension with TCS, and that after a period of familiarisation listeners report being uncomfortable with speech presented at normal rates.

This paper describes recent developments in speeded listening and associated devices, and discusses applications in CBL.