From: Sentiment analysis classification system using hybrid BERT models
Model | Dataset | Accuracy% | Notes |
---|---|---|---|
DistilBERT with emojis | Airlines | 83.74 | GLG |
Crowdflower | 80.42 | GLG | |
Apple | 86.81 | LGL | |
DistilBERT without emojis | Airlines | 83.47 | GLG |
Crowdflower | 79.24 | GLG | |
Apple | 88.04 | GLG | |
RoBERTa with emojis | Airlines | 86 | 3G |
Crowdflower | 82.39 | – | |
Apple | 91.72 | – | |
RoBERTa without emojis | Airlines | 85.93 | GLG |
Crowdflower | 81.34 | 3L | |
Apple | 91.72 | 3G | |
Indrayuni et al. [49] | Apple products | 85.76 | SVM + GA |
Dang et al. [50] (two classes) | Sentiment140 | 80 | Word embeeding-RNN |
Tweets SemEval | 85 | Word embeeding-RNN | |
IMDB Movie Reviews (1) | 87 | Word embeeding-RNN | |
IMDB Movie Reviews (2) | 86 | Word embeeding-RNN | |
Cornell Movie Reviews | 76 | Word embeeding-RNN | |
Book Reviews | 76 | Word embeeding-CNN | |
Music Reviews | 76 | TF-IDF-DNN | |
Tweets Airline | 90 | Word embeeding-RNN | |
Kumawat et al. [51] | Twitter US Airline Sentiment | 81.2 | BERT |
80.8 | RoBERTa | ||
Xiang [52] | Twitter collection | 76.6 | BiLSTM(EPA) |
airline dataset | 82 | BiLSTM(P) | |
IMDB review | 82.6 | BiLSTM-AT(P) | |
Shuang [53] | Twitter airlines | 83.3 | M_ARC |
Yelp | 79.1 | RC | |
Janjua et al. [54] | Sanders Twitter Corpus (STC) | 78.99 | MuLeHyABSC + MLP |
Twitter Airline Sentiment (TAS) | 84.09 | ||
First GOP Debate (FGD) | 80.38 | ||
Apple Twitter Corpus (ATC) | 82.37 | ||
Stanford Twitter Sentiment (STS) | 84.72 | ||
Kian [43] | Twitter US Airline Sentiment | 85.89 | RoBERTa-LSTM |
Twitter US Airline Sentiment Augmented | 91.37 | ||
IMDB | 92.96 | ||
Sentiment140 | 89.70 | ||
Barakat [45] | Airline | 99.78 | ULMFit-SVM |
Jain [44] (two classes) | Airlinequality Airline Sentiment Data | 90.2 | CNN-LSTM |
Twitter Airline Sentiment Data | 91.3 | ||
Thapa et al. [55] | 60 | VADER | |
70 | |||
Demotte et al. [56] | CrowdFlower US Airline | 82.04 | GloVe + shallow capsule network with static routing |
Twitter Sentiment Gold | 86.87 |