Preprint
Kid-Whisper: Towards Bridging the Performance Gap in Automatic Speech Recognition for Children VS. Adults
Ahmed Adel Attia et al., arXiv (eess.AS), revised May 2024, 2023
Read the originalRead on October 16, 2026
- What it is
- A study that adapts OpenAI's Whisper speech recognition model to children's speech, reporting word error rates on the MyST children's classroom speech test set before and after better data preprocessing.
- How much weight it carries
- A peer-reviewed-style preprint from university speech researchers with public code. It measures research models on school-age children's speech in a few corpora, not consumer dictation apps, so its numbers show the size of the problem rather than the accuracy of any product.
Facts taken from this source
- The Kid-Whisper study reports reducing the word error rate of Whisper-Small on the MyST children's test set from 13.93% to 9.11%.
“We reduce the Word Error Rate (WER) on the MyST testset 13.93% to 9.11% with Whisper-Small” (Abstract)