Preprint
Causal Analysis of ASR Errors for Children: Quantifying the Impact of Physiological, Cognitive, and Extrinsic Factors
Vishwanath Pratap Singh et al., arXiv, 2025
Read the originalRead on October 16, 2026
- What it is
- A 2025 preprint that benchmarks four speech foundation models, Wav2Vec2.0, HuBERT, Whisper and MMS, on two children's speech corpora, grouped by age, then uses causal inference to rank the factors behind their errors.
- How much weight it carries
- A preprint with a clear method and two public children's corpora. It covers read and tutoring-style speech and older model versions, so the exact error rates will differ for current apps, and it has not been through peer review.
Facts taken from this source
- In a 2025 study of speech recognition models on children's speech, a child's age and the number of words in the audio had the highest impact on accuracy, followed by background noise and pronunciation ability.
“physiology (age) and particular external factor (number of words in audio) have the highest impact on accuracy, followed by background noise and pronunciation ability” (Abstract)