Peer-reviewed research
Longitudinal study of ASR performance on ageing voices
Ravichander Vipperla et al., Interspeech 2008 (ISCA), 2008
Read the originalRead on October 17, 2026
- What it is
- A longitudinal study from the University of Edinburgh that ran a speech recognizer on recordings of the US Supreme Court over several years. It compared the word error rate for elderly voices with the rate for adult voices.
- How much weight it carries
- Peer-reviewed and accepted after full review at Interspeech 2008. It uses one court corpus and recognizers from 2008, which are much weaker than current models, so the size of the gap is out of date even though the direction is informative.
Facts taken from this source
- In a 2008 longitudinal study on court recordings, the word error rate of a speech recognizer rose gradually as the age of elderly speakers increased.
“The word error rate increases gradually as the age of the elderly speakers increase” (Abstract)