Application of machine learning techniques to automate speech mispronunciation detection in children and adolescents
Nazila Ameli, Mahshid Nik Ravesh, Luan Matheus Trindade Dalmazo, Karen Pollock, Daniel DeSantis, Manuel Lagravere, Hollis Lai · 2026
Read Application of machine learning techniques to automate speech mispronunciation detection in children and adolescents by Nazila Ameli, Mahshid Nik Ravesh, Luan Matheus Trindade Dalmazo, Karen Pollock, Daniel DeSantis, Manuel Lagravere, Hollis Lai (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13668468.txt.
Description
Nazila Ameli and six co-authors train a deep learning system to flag mispronounced sounds in children's speech recordings. Speech sound disorders are common and can affect school and social outcomes, but expert listening assessment is slow and varies between raters. Automated screening could triage children where specialists are scarce.
One hundred participants aged six to eighteen recorded eighteen hundred words under a protocol focused on fricative sounds. Two expert speech-language pathologists labeled each recording, with a third settling disagreements, and reliability was measured carefully. A pretrained transformer speech model was fine-tuned on the cleaned audio for present-or-absent classification. Imbalance tricks included augmentation, weighted sampling, and focal loss. Performance was judged on a held-out test set.
Catalog record
| CREATOR | Nazila Ameli, Mahshid Nik Ravesh, Luan Matheus Trindade Dalmazo, Karen Pollock, Daniel DeSantis, Manuel Lagravere, Hollis Lai |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |