parakeet-tdt-0.6b-lt-study-s-d80-dial

Research checkpoint from the study Lithuanian speech recognition: effects of dialect training and transcript spelling. Fine-tuned from nvidia/parakeet-tdt-0.6b-v3 on LIEPA-3 telephone speech (420.17 h) + LIEPA-3 dialect speech (82.30 h), dialect clips in dialect spelling. Output is lowercase spoken form without punctuation. This is a study model, not a product model.

WER (%)

Test	WER
Dialect test, dialect-spelling reference	31.28
Dialect test, standard-spelling reference	36.29
Dialect test, either spelling accepted	27.60
LIEPA-3 telephone test	10.80
FLEURS lt test	23.24
Common Voice 19 lt test	25.77

Use

import nemo.collections.asr as nemo_asr
model = nemo_asr.models.ASRModel.from_pretrained("Digisensus/parakeet-tdt-0.6b-lt-study-s-d80-dial")
print(model.transcribe(["audio.wav"])[0].text)

Provenance

Run	ltd26/S+D80.dial/s1/a3
Recipe	S+D80.dial (recipes/S+D80.dial.json in the code repo)
Splits	splits-v1
Training	10,000 updates, 192 clips per update, last checkpoint; full spec in config.json
Scorer	see registry/scores.csv in the code repo