BabyHuBERT, a self‑supervised model for child recordings, was trained on over 13,000 hours of audio in 40 languages and achieved F1 scores of 52.1%‑74.4%, beating HuBERT. Read more: https://getnews.me/babyhubert-multilingual-ai-model-improves-child-speech-segmentation/ #babyhubert #speechmodel
September 20, 2025 at 6:03 AM
BabyHuBERT, a self‑supervised model for child recordings, was trained on over 13,000 hours of audio in 40 languages and achieved F1 scores of 52.1%‑74.4%, beating HuBERT. Read more: https://getnews.me/babyhubert-multilingual-ai-model-improves-child-speech-segmentation/ #babyhubert #speechmodel
Charlot, Kunze, Poli, Cristia, Dupoux, Lavechin: BabyHuBERT: Multilingual Self-Supervised Learning for Segmenting Speakers in Child-Centered Long-Form Recordings https://arxiv.org/abs/2509.15001 https://arxiv.org/pdf/2509.15001 https://arxiv.org/html/2509.15001
September 19, 2025 at 6:36 AM
Charlot, Kunze, Poli, Cristia, Dupoux, Lavechin: BabyHuBERT: Multilingual Self-Supervised Learning for Segmenting Speakers in Child-Centered Long-Form Recordings https://arxiv.org/abs/2509.15001 https://arxiv.org/pdf/2509.15001 https://arxiv.org/html/2509.15001