List your company
ResearchFormerly WizWand

Intent Classification on SLURP (test)

90.07Accuracy (IC)TDT 0-8
0.432423.703746.97570.2463Nov 4, 2021Jan 30, 2022Apr 28, 2022Jul 24, 2022Oct 20, 2022Jan 15, 2023Apr 13, 2023
Updated 8mo ago

Evaluation results

MethodLinks
TDT 0-8#params (M)=119, relative speedup=1.28x, max duration=82023.0490.07
TDT 0-4#params (M)=119, relative speedup=1.17x, max duration=42023.0489.85
EF-hbt-largeFine-tuning strategy=Entire, Encoder Backbone=HuBERT, Model Size=Large2021.1189.38
TDT 0-6#params (M)=119, relative speedup=1.28x, max duration=62023.0489.28
PF-hbt-largeFine-tuning strategy=Partial, Encoder Backbone=HuBERT, Model Size=Large2021.1189.22
EF-hbt-large-960hFine-tuning strategy=Entire, Encoder Backbone=HuBERT, Model Size=Large, Pre-training Corpus=960h2021.1188.71
RNNT#params (M)=119, relative speedup=-2023.0488.53
PF-hbt-large-960hFine-tuning strategy=Partial, Encoder Backbone=HuBERT, Model Size=Large, Pre-training Corpus=960h2021.1188.32
SpeechBrain#params (M)=96, relative speedup=N/A2023.0487.7
PF-hbt-baseFine-tuning strategy=Partial, Encoder Backbone=HuBERT, Model Size=Base2021.1187.51
EF-hbt-baseFine-tuning strategy=Entire, Encoder Backbone=HuBERT, Model Size=Base2021.1187.44
EF-w2v-baseFine-tuning strategy=Entire, Encoder Backbone=wav2vec 2.0, Model Size=Base2021.1187.13
TDT 0-2#params (M)=119, relative speedup=1.17x, max duration=22023.0487.12
CTI2021.1186.92
PF-w2v-baseFine-tuning strategy=Partial, Encoder Backbone=wav2vec 2.0, Model Size=Base2021.1186.58
ESPnet-SLU#params (M)=109, relative speedup=N/A2023.0486.52
PF-w2v-large-960hFine-tuning strategy=Partial, Encoder Backbone=wav2vec 2.0, Model Size=Large, Pre-training Corpus=960h2021.1186.35
PF-w2v-largeFine-tuning strategy=Partial, Encoder Backbone=wav2vec 2.0, Model Size=Large2021.1186.29
PF-w2v-base-960hFine-tuning strategy=Partial, Encoder Backbone=wav2vec 2.0, Model Size=Base, Pre-training Corpus=960h2021.1186.13
EF-w2v-large-960hFine-tuning strategy=Entire, Encoder Backbone=wav2vec 2.0, Model Size=Large, Pre-training Corpus=960h2021.1186.1
EF-w2v-base-960hFine-tuning strategy=Entire, Encoder Backbone=wav2vec 2.0, Model Size=Base, Pre-training Corpus=960h2021.1185.89
EF-w2v-largeFine-tuning strategy=Entire, Encoder Backbone=wav2vec 2.0, Model Size=Large2021.1185.8
Frozen-hbt-largeFine-tuning strategy=Frozen, Encoder Backbone=HuBERT, Model Size=Large2021.1174.42
Frozen-hbt-baseFine-tuning strategy=Frozen, Encoder Backbone=HuBERT, Model Size=Base2021.1168.74
Frozen-w2v-baseFine-tuning strategy=Frozen, Encoder Backbone=wav2vec 2.0, Model Size=Base2021.1147.15
Frozen-w2v-largeFine-tuning strategy=Frozen, Encoder Backbone=wav2vec 2.0, Model Size=Large2021.113.88