espnet/proyecto_nahuatl

ESPnet2 ASR model

espnet/proyecto_nahuatl

This model was trained using proyecto_nahuatl recipe in espnet.

Demo: How to use in ESPnet2

cd espnet/egs2/proyecto_nahuatl
# add data path to prefix in run.sh
./run.sh 

RESULTS

Environments

• date: Tue May 6 10:18:30 EDT 2025
• python version: 3.9.21 (main, Dec 11 2024, 16:24:11) [GCC 11.2.0]
• espnet version: espnet 202412
• pytorch version: pytorch 2.2.0+cu121
• Git hash: 90ea0823e6d0aa2ea7b322a8b513cbb0b8200e91
  • Commit date: Mon May 5 20:04:01 2025 +0000

exp/asr_train_asr_s3prl_raw_en_bpe150_sp/decode_asr_ctc_lm_lm_train_lm_en_bpe150_valid.loss.ave_asr_model_valid.loss.ave

WER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
test/Hidalgo	324	3760	9.9	57.4	32.7	8.0	98.1	100.0
test/Tequila	870	9337	14.0	62.5	23.4	8.8	94.8	100.0
test/Zacatlan	1836	16318	21.9	56.6	21.5	5.1	83.2	99.6

CER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
test/Hidalgo	324	23998	62.4	13.0	24.6	15.3	52.9	100.0
test/Tequila	870	65418	70.4	11.6	18.0	14.2	43.8	100.0
test/Zacatlan	1836	108155	79.9	7.6	12.5	9.4	29.5	99.6

TER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
test/Hidalgo	324	13202	48.2	27.2	24.6	15.9	67.7	100.0
test/Tequila	870	37224	55.5	25.3	19.3	13.8	58.3	100.0
test/Zacatlan	1836	61871	66.2	19.4	14.4	8.4	42.2	99.6