Phone Recognition on TIMIT (test)

17.3Frame Error Rate

Segmental RNN

Updated 1mo ago

Evaluation Results

Method
Segmental RNN 2016.03	17.3	-	-	-
Attention-based RNN 2016.03	17.6	-	-	-
RNN transducer 2016.03	17.7	-	-	-
CTC 2016.03	18.4	-	-	-
HMM-DNN 2016.03	18.5	-	-	-
Segmental RNN 2016.03	18.9	-	-	-
Discriminative segmental cascade + 2nd pass with various features 2016.03	19.9	-	-	-
Discriminative segmental cascade 2016.03	21.7	-	-	-
Deep Segmental NN 2016.03	21.9	-	-	-
Boundary-factored SCRF 2016.03	26.5	-	-	-
Bidirectional LSTM 2015.04	29.1	-	-	-
Bidirectional iRNN 2015.04	29.7	-	-	-
Bidirectional LSTM 2015.04	30.6	-	-	-
Bidirectional RNN 2015.04	32.4	-	-	-
first-pass SCRF 2016.03	33.1	-	-	-
Bidirectional iRNN 2015.04	33.2	-	-	-
iRNN 2015.04	33.8	-	-	-
Bidirectional RNN 2015.04	34.8	-	-	-
LSTM 2015.04	35.4	-	-	-
iRNN 2015.04	35.5	-	-	-
RNN 2015.04	36.2	-	-	-
LSTM 2015.04	36.2	-	-	-
RNN 2015.04	37	-	-	-
RecurrentAreas with ABS plasticity 2026.03	-	47.5	2.6	-
Espeak G2P Baseline 2026.06	-	-	-	4.97
Allosaurus 2026.06	-	-	-	10
W2V2-eSpeak 2026.06	-	-	-	5.9
W2V2-Vitouphy 2026.06	-	-	-	6
MultIPA 2026.06	-	-	-	12.3
HuBERT Phoneme 2026.06	-	-	-	6.4
POWSM 2026.06	-	-	-	9.3
WavLM HuPER 2026.06	-	-	-	6.7
Optimal Curriculum 2026.06	-	-	-	2.1