datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
viet_muong_vtv5-201-300-full-preprocessingViMD_preprocessing
ViMD Truncated 10s — 16kHz
Preprocessed version of ViMD (Nguyen et al., EMNLP 2024) for Dialect Identification.
Preprocessing applied
Resample: 44.1kHz (original) → 16kHz, mono
Truncate: only the FIRST 10 SECONDS of each audio are kept
(files shorter than 10s are kept intact). 1 original file = 1 sample.
This follows the truncation strategy of Lu et al. (2020), NOT chunking.
Splits: original ViMD train/valid/test kept unchanged (speaker-exclusive).… See the full description on the dataset page: https://huggingface.co/datasets/tannhoo06/ViMD_preprocessing.