RESULTS.md

metadata

tags:
  - espnet
  - audio
  - speech-recognition
language: en
datasets:
  - google/fleurs
license: cc-by-4.0

ESPnet2 ASR model

espnet/wanchichen_fleurs_multilingual_asr_hubert_frontend

This model was trained by William Chen using the fleurs recipe in espnet.

Demo: How to use in ESPnet2

cd espnet
pip install -e .
cd egs2/fleurs/asr1
./run.sh

RESULTS

Environments

• date: Sun Aug 21 15:18:30 EDT 2022
• python version: 3.8.6 (default, Dec 17 2020, 16:57:01) [GCC 10.2.0]
• espnet version: espnet 202205
• pytorch version: pytorch 1.8.1+cu102
• Git hash: 45e8cb9173a072f85ee7a7ccbcae06af7c5c484a
  • Commit date: Wed Jun 1 14:21:14 2022 +0900

asr_train_asr_hubert_raw_all_raw_all_bpe6500_sp

WER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
decode_asr_asr_model_valid.acc.best/test_all	77809	1669969	54.2	38.1	7.8	4.0	49.9	99.7
decode_asr_asr_model_valid.acc.best/dev_all	31622	610500	53.4	38.7	7.9	3.5	50.1	99.5

CER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
decode_asr_asr_model_valid.acc.best/test_all	77809	10876592	85.1	7.4	7.5	3.2	18.0	99.7
decode_asr_asr_model_valid.acc.best/dev_all	31622	3988181	84.7	7.7	7.7	2.8	18.1	99.5

TER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
decode_asr_asr_model_valid.acc.best/test_all	77809	9622352	82.8	9.2	8.0	3.4	20.6	99.7
decode_asr_asr_model_valid.acc.best/dev_all	31622	3547834	82.4	9.4	8.2	3.0	20.6	99.5