List your company
ResearchFormerly WizWand

Sentiment Classification on Yelp Polarity (test)

1.81Error RateBERT-ITPT-FiT
1.30324.72418.14511.5659Sep 4, 2015Apr 15, 2016Nov 26, 2016Jul 9, 2017Feb 18, 2018Oct 1, 2018May 14, 2019
Updated 8mo ago

Evaluation results

MethodLinks
BERT-ITPT-FiTBackbone=BERT-Large, Further pre-training=In-Task (ITPT)2019.051.81
BERT-ITPT-FiTBackbone=BERT-Base, Further pre-training=In-Task (ITPT)2019.051.92
BERT-FiTBackbone=BERT-Large2019.052.04
ULMFiT2019.052.16
BERT-FiTBackbone=BERT-Base2019.052.28
M-ACNNNumber of convolutional filters=Multiple2017.093.89
Self-attentive Embedding2017.093.92
VDCNNDepth=29, Pooling=MaxPooling2016.064.28
Deep CNNNetwork depth=29 layers2017.094.28
VDCNNDepth=29, Pooling=Convolution2016.064.35
ngrams2015.094.36
ngramsAuthor=Zhang2016.064.36
ngramsType=bag-of-means, N-grams=up to 5-grams2017.094.36
VDCNNDepth=17, Pooling=MaxPooling2016.064.5
Deep CNNNetwork depth=17 layers2017.094.5
ngrams TFIDFTF-IDF Weighting=true2015.094.56
ngrams TFIDFType=TFIDF, N-grams=up to 5-grams2017.094.56
M-CNNNumber of convolutional filters=Multiple2017.094.58
VDCNNDepth=29, Pooling=KMaxPooling2016.064.63
Full Conv. (large, Th)Model Size=large, Case Sensitive=true, Thesaurus Augmentation=true2015.094.88
VDCNNDepth=9, Pooling=MaxPooling2016.064.88
Deep CNNNetwork depth=9 layers2017.094.88
Large word CNNScale=Large, Features per layer=1024, Total layers=62017.094.89
VDCNNDepth=17, Pooling=Convolution2016.064.96
VDCNNDepth=9, Pooling=Convolution2016.065.01
Lk. Conv. (large, Th)Model Size=large, Thesaurus Augmentation=true2015.095.03
VDCNNDepth=17, Pooling=KMaxPooling2016.065.05
LSTM2015.095.26
VDCNNDepth=9, Pooling=KMaxPooling2016.065.27
Full Conv. (small, Th)Model Size=small, Case Sensitive=true, Thesaurus Augmentation=true2015.095.42
Small word CNNScale=Small, Features per layer=256, Total layers=62017.095.54
Conv. (large, Th)Model Size=large, Thesaurus Augmentation=true2015.095.82
BoW TFIDFTF-IDF Weighting=true2015.096.34
S-ACNNNumber of convolutional filters=Single2017.096.41
BoW2015.097.76
Bag-of-means2015.0912.67
S-CNNNumber of convolutional filters=Single2017.0914.48