Table 6 Summary of the proposed models in the state-of-the-art tweet sentiment analysis literature.
Study | Model used | Dataset | Macro average precision | Macro average recall | Macro average F1-score | Accuracy |
|---|---|---|---|---|---|---|
Qi & Shabrina (2023)16 | BoW+SVC | UK COVID-19 Twitter Dataset | 69.66% | 70.33% | 69.66% | 71.00% |
Ours | TRABSA | UK COVID-19 Twitter Dataset | 94.00% | 93.00% | 94.00% | 94.00% |
dos Santos Neto et al., (2024)49 | BERT | TripAdvisor | 87.70% | 88.20% | 87.90% | 88.20% |
Brum & Volpe Nunes (2018)50 | BERT | TweetSentBR | 73.27% | 72.75% | 72.96% | 72.75% |
De Souza et al. (2018)51 | MultiFiT-Twitter LM | Twitter NPS | 72.43% | 72.46% | 72.43% | 72.46% |
Pilar et al. (2023)52 | Neighbor-sentiment | InterTASS | 57.76% | 51.39% | 54.39% | 61.35% |
Su & Kabala (2023)53 | GloVe100+LSTM | 500k ChatGPT-related Tweets Jan-Mar 2023 | 81.10% | 81.10% | 81.10% | 81.10% |
Memiş et al. (2024)54 | Multiclass CNN model with pre-trained word embedding | Turkish Financial Tweets | – | – | – | 72.73% |
Kp et al. (2024)55 | Ensemble classifier | Twitter API Dataset | 91.29% | 89.65% | 87.32% | 93.42% |
Mohbey et al. (2024)56 | CNN-LSTM | Monkeypox Tweets | 91.24% | 91.24% | 91.24% | 91.24% |
Sazan et al., 2024)57 | RoBERTa+fastText | US Airline Dataset | 92.08% | 92.02% | 92.05% | 92.02% |
Ours | TRABSA | US Airline Dataset | 95.00% | 95.00% | 95.00% | 96.00% |
Jlifi et al. (2024)58 | Ens-RF-BERT | Hashtag Covid19 Tweets | 94.03% | 93.05% | 94.03% | 93.01% |
Bhardwaj et al. (2024)59 | BoW+LR | COV19Tweets Dataset | 82.00% | 81.80% | 81.60% | 81.80% |
Ours | TRABSA | Global COVID-19 dataset | 98.00% | 98.00% | 98.00% | 98.00% |
