List your company
ResearchFormerly WizWand

Sentiment Classification on CR

91.4AccuracyBYTE MLSTM
80.469683.307386.14588.9827Aug 25, 2014Jul 13, 2016Jun 2, 2018Apr 21, 2020Mar 11, 2022Jan 29, 2024Dec 18, 2025
Updated 8mo ago

Evaluation results

MethodLinks
BYTE MLSTMClassifier=Logistic Regression, Penalty=L12017.0491.4
byte mLSTMd*=40962018.0590.6
SBERT2025.1290
CNN+MCFAAdditional Context=Translation, N=ensemble2018.0689.4
RoBERTa (Gao et al., 2021)# Params=1.0x, Evaluation Protocol=Full fine-tuning2022.1289.4
CNN+MCFAAdditional Context=Translation, N=102018.0688.6
SCD-RoBERTa_baseBackbone=RoBERTa-base2022.0387.76
CNN+MCFAAdditional Context=Translation, N=12018.0687.6
USE_T+CNN (w2v w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=CNN2018.0387.45
USE TUniversal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=None, Transfer Task Model=None2018.0387.43
BERT [CLS]-embeddingBackbone=BERT-base, Pooling=CLS2022.0387.39
SimCSE-RoBERTa_baseBackbone=RoBERTa-base2022.0387.28
IS-BERT_baseBackbone=BERT-base2022.0387.18
CNN+B1Additional Context=Translation, N=ensemble2018.0686.7
USE_T+DAN (w2v w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=DAN2018.0386.66
CNN+B2Additional Context=Translation, N=ensemble2018.0686.6
USE_T+CNN (lrn w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=CNN2018.0386.45
TopCNNAdditional Context=Topic, Granularity=ensemble2018.0686.4
AdaSent2018.0686.3
CNN+B2Additional Context=Translation, N=102018.0686.3
BiLSTM-Max (on AllNLI)Training Methodology=Supervised, Source Dataset=AllNLI2017.0586.3
AdaSentTraining Methodology=Supervised, Transfer Protocol=No Transfer2017.0586.3
AdaSent2015.0486.3
ADASENT2017.0486.3
Avg. BERT embeddingsBackbone=BERT-base, Pooling=Avg2022.0386.25
CNN+B1Additional Context=Translation, N=12018.0686.2
CNN+B2Additional Context=Translation, N=12018.0686.1
USE_T+DAN (lrn w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=DAN2018.0386.08
MC-QTd*=48002018.0586
CNN+B1Additional Context=Translation, N=102018.0685.9
SCD-BERT_baseBackbone=BERT-base2022.0385.8
SimCSE-BERT_baseBackbone=BERT-base2022.0385.75
CNN-Rule-qNetwork role=teacher network, Rule type=but-rule2016.0385.3
Capsule-B2018.0385.1
CNN-multichannelword vectors=pre-trained (word2vec), channels=22014.0885
CNN2018.0685
CNN2015.0485
CNN2017.0485
CNN-Rule-pNetwork role=student network, Rule type=but-rule2016.0385
CNN-multichannel2016.0385
TopCNNAdditional Context=Topic, Granularity=word2018.0684.9
TopCNNAdditional Context=Topic, Granularity=sentence2018.0684.8
RoBERTa [CLS]-embeddingBackbone=RoBERTa-base, Pooling=CLS2022.0384.77
CNN-staticword vectors=static pre-trained (word2vec)2014.0884.7
CNNmode=static2018.0384.7
CNN-non-staticword vectors=fine-tuned pre-trained (word2vec)2014.0884.3
à la carten=3, d*=48002018.0584.3
CNNmode=non-static2018.0384.3
CNNModel variant=non-static2016.0384.3
GPT-2 kNN-LM# Params=2.2x, Evaluation Protocol=Zero-shot2022.1284.3
CNN-400Emb.Size=4002019.0583.82
NLM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=CNN2019.0583.82
skip-thoughtsd*=48002018.0583.8
Capsule-A2018.0383.8
à la carten=2, d*=32002018.0583.7
SCLie2025.1283.6
STM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=CNN2019.0583.57
LM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=LSTM2019.0583.54
LM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=CNN2019.0583.4
NLM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=LSTM2019.0583.32
Tree-LSTM2018.0383.2
SkipThought-LNTraining Methodology=Unsupervised, Sentence Ordering=Ordered, Transfer Protocol=Transfer2017.0583.1
SKIPTHOUGHTLayer Normalization=True2017.0483.1
NLM+TSEDEmb.Size=50, Teacher Model Architecture=CNN2019.0583.04
LSTM-400Emb.Size=4002019.0582.98
NLM+TSEDEmb.Size=50, Teacher Model Architecture=LSTM2019.0582.96
STM+TSEDEmb.Size=50, Teacher Model Architecture=LSTM2019.0582.96
GPT-2 + PMI# Params=2.2x, Evaluation Protocol=Zero-shot2022.1282.8
LM+TSEDEmb.Size=50, Teacher Model Architecture=LSTM2019.0582.76
STM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=LSTM2019.0582.73
CRF-PR2014.0882.7
CRF-PR2016.0382.7
BRNN2015.0482.6
DPCLie2025.1282.6
LR-LSTM2018.0382.5
DPCNN2025.1282.4
RNN2015.0482.3
G-Dropout2014.0882.1
BILSTM2018.0382.1
G-Dropout2016.0382.1
USE_D+CNN (w2v w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=CNN2018.0382.04
CNN-LSTMd*=48002018.0582
USE_D+DAN (lrn w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=DAN2018.0381.93
F-Dropout2014.0881.9
LM+TSEDEmb.Size=50, Teacher Model Architecture=CNN2019.0581.84
NBSVM2014.0881.8
NB-SVM2015.0481.8
NBSVM2017.0481.8
SKIPTHOUGHT2017.0481.8
LinearArchitecture=Fully connected Neural Network replacement for Convolution layer2025.1281.8
USE_D+DAN (w2v w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=DAN2018.0381.71
DisCn=2-3, d*=3200-48002018.0581.5
SCNN2025.1281.5
USE_D+CNN (lrn w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=CNN2018.0381.49
Tree-CRF2014.0881.4
à la carten=1, d*=16002018.0581.3
GrConv2015.0481.3
NPM# Params=1.0x, Evaluation Protocol=Zero-shot, nonparametric=true2022.1281.2
USE DUniversal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence, Embedding Initialization=None, Transfer Task Model=None2018.0380.97
CNN-50Emb.Size=502019.0580.89