Table 1 Comparison of sentiment analysis models on user reviews.
Paper/study | Dataset used | Model(s) applied | Accuracy | F1-Score | Notes |
|---|---|---|---|---|---|
Samanmali et al.57 | Google Play App Reviews of 15 popular Apps | Logistic Regression, Naïve Bayes, SVM | \(\sim\) 92.75% | 96.3% | LSTM outperformed ANN and SVM with the highest accuracy |
González et al.58 | IMDB movie reviews | BERT, Logistic Regression, SVM, Naïve Bayes | 93.87% | Not provided | BERT outperformed all models |
Eser & Sahin59 | Spotify App Reviews (Google Play Store) | BERT, DistilBERT, RoBERTa, XLM-RoBERTa | 71.68 (DistilBERT) | 69.24 (XLM-RoBERTa) | DistilBERT achieved highest accuracy; XLM-RoBERTa had best F1 score |
Perikos & Diamantopoulos60 | Naver Dataset | RoBERTa, DistilBERT, XLNet | RoBERTa-large (97.62%) | RoBERTa-large (94.77%) | RoBERTa-large outperformed other models, showing superior accuracy and F1 score |