Chaitanya Narisetty
transducer v2 model files
f173429

RESULTS

Environments

  • date: Tue Apr 26 15:33:18 EDT 2022
  • python version: 3.9.5 (default, Jun 4 2021, 12:28:51) [GCC 7.5.0]
  • espnet version: espnet 202204
  • pytorch version: pytorch 1.8.1+cu111
  • Git hash: 8a76ff24eb513d96561fb47d0320dd39c1c3645a
    • Commit date: Tue Apr 19 07:32:58 2022 +0000

asr_train_conformer-rnn_transducer_raw_en_bpe5000_sp

WER

dataset Snt Wrd Corr Sub Del Ins Err S.Err
decode_asr_model_valid.loss.ave_10best/dev_clean 2703 54402 97.7 2.1 0.2 0.3 2.6 31.5
decode_asr_model_valid.loss.ave_10best/dev_other 2864 50948 93.8 5.6 0.6 0.6 6.8 50.8
decode_asr_model_valid.loss.ave_10best/test_clean 2620 52576 97.5 2.3 0.2 0.3 2.8 32.7
decode_asr_model_valid.loss.ave_10best/test_other 2939 52343 94.1 5.3 0.6 0.7 6.6 51.8
decode_lm_weight0.3_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 52576 97.8 2.0 0.2 0.3 2.5 29.4
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/dev_clean 2703 54402 98.0 1.8 0.2 0.2 2.2 28.2
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/dev_other 2864 50948 94.8 4.5 0.7 0.5 5.7 45.1
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 52576 97.9 1.9 0.2 0.3 2.4 29.3
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_other 2939 52343 94.9 4.3 0.7 0.5 5.6 47.0
decode_lm_weight0.5_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 52576 97.8 1.9 0.3 0.3 2.4 29.5

CER

dataset Snt Wrd Corr Sub Del Ins Err S.Err
decode_asr_model_valid.loss.ave_10best/dev_clean 2703 288456 99.4 0.4 0.3 0.2 0.9 31.5
decode_asr_model_valid.loss.ave_10best/dev_other 2864 265951 97.7 1.4 0.9 0.8 3.0 50.8
decode_asr_model_valid.loss.ave_10best/test_clean 2620 281530 99.4 0.4 0.3 0.3 0.9 32.7
decode_asr_model_valid.loss.ave_10best/test_other 2939 272758 97.9 1.2 0.9 0.8 2.8 51.8
decode_lm_weight0.3_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 281530 99.4 0.3 0.3 0.2 0.8 29.4
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/dev_clean 2703 288456 99.4 0.3 0.3 0.2 0.8 28.2
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/dev_other 2864 265951 97.9 1.1 1.0 0.6 2.7 45.1
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 281530 99.4 0.3 0.3 0.2 0.9 29.3
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_other 2939 272758 98.1 0.9 1.0 0.6 2.5 47.0
decode_lm_weight0.5_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 281530 99.3 0.3 0.4 0.2 0.9 29.5

TER

dataset Snt Wrd Corr Sub Del Ins Err S.Err
decode_asr_model_valid.loss.ave_10best/dev_clean 2703 68010 97.2 2.1 0.7 0.4 3.3 31.5
decode_asr_model_valid.loss.ave_10best/dev_other 2864 63110 92.7 5.6 1.7 1.2 8.6 50.8
decode_asr_model_valid.loss.ave_10best/test_clean 2620 65818 97.0 2.2 0.9 0.4 3.4 32.7
decode_asr_model_valid.loss.ave_10best/test_other 2939 65101 93.0 5.1 1.9 1.0 8.0 51.8
decode_lm_weight0.3_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 65818 97.3 1.9 0.8 0.4 3.1 29.4
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/dev_clean 2703 68010 97.5 1.8 0.8 0.4 2.9 28.2
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/dev_other 2864 63110 93.5 4.5 1.9 0.9 7.4 45.1
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 65818 97.3 1.9 0.8 0.4 3.0 29.3
decode_lm_weight0.4_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_other 2939 65101 93.9 4.1 1.9 0.8 6.9 47.0
decode_lm_weight0.5_lm_lm_train_lm_transformer2_en_bpe5000_17epoch_asr_model_valid.loss.ave_10best/test_clean 2620 65818 97.3 1.9 0.9 0.4 3.1 29.5