Automatic Speech Recognition
audio
vosk
kaldi
nextcloud
kyteinsky's picture
initial commit
22028d0 verified
Esperanto small model for Vosk
WER
%WER 4.80 [ 5824 / 121229, 809 ins, 549 del, 4466 sub ] exp/chain/tdnn/decode_dev_look_fast/wer_11_0.0
%WER 7.24 [ 8673 / 119712, 1073 ins, 1060 del, 6540 sub ] exp/chain/tdnn/decode_test_look_fast/wer_11_0.0
Previous model 0.22
%WER 6.02 [ 7294 / 121229, 931 ins, 671 del, 5692 sub ] exp/chain.save1/tdnn/decode_dev_look_fast/wer_11_0.0
%WER 8.55 [ 10238 / 119712, 1191 ins, 1232 del, 7815 sub ] exp/chain.save1/tdnn/decode_test_look_fast/wer_11_0.0