Harcuracy/google_waxal_asr_challenge

WaxalNLP ASR — Cleaned Subset (Lingala, Shona, Luganda) This dataset is a cleaned, corrected subset of google/WaxalNLP, covering the train and validation splits for three languages: lin_asr — Lingala sna_asr — Shona lug_asr — Luganda The test split from the original dataset is intentionally excluded. What was changed The original transcriptions for these three languages contained a number of errors. A corrected transcription file was applied on top of the… See the full description on the dataset page: https://huggingface.co/datasets/Harcuracy/google_waxal_asr_challenge.