MyST

The MyST dataset is a child speech corpus used to evaluate the performance of speech models in low-resource domains.

Papers using MyST (9)

Gumbel-BEARD: Automatic Layer Selection for Self-Supervised Adaptation of Whisper in Low-Resource Domains2026Selective Attention Merging for low resource tasks: A case study of Child ASR2025 · 2 citesG-IFT: A Gated Linear Unit adapter with Iterative Fine-Tuning for Low-Resource Children's Speaker Verification2025K-Function: Joint Pronunciation Transcription and Feedback for Evaluating Kids Language Function2025A Wav2vec2-Based Experimental Study on Self-Supervised Learning Methods to Improve Child Speech Recognition2022 · 5 citesImproved Child Text-to-Speech Synthesis through Fastpitch-based Transfer Learning2023 · 2 citesUniEnc-CASSNAT: An Encoder-only Non-autoregressive ASR for Speech SSL Models2024 · 2 citesDRAFT: A Novel Framework to Reduce Domain Shifting in Self-supervised Learning and Its Application to Children's ASR2022Towards Better Domain Adaptation for Self-supervised Models: A Case Study of Child ASR2023