whyismydininghallonfire/

orig-plus-asr-tamil-clean

Combined ASR dataset built from:

albagon/til26-asr-split (orig rows) whyismydininghallonfire/asr-tamil-clean (asr_tamil_clean rows)

Audio paths are namespaced under each split to avoid filename collisions:

audio/orig/... audio/asr_tamil_clean/...

Each row keeps key, audio, transcript, and language, with an added source_dataset field. Counts:

train: 3595 orig + 891 asr_tamil_clean = 4486 validation: 899 orig + 224 asr_tamil_clean = 1123