| Lexicon based approach |
Manual or Predefined Lexicons |
Product reviews, Social media |
Naive Bayes, Lexicon based |
Performance varies |
Liu & Shen (2020), Barik & Misra (2024)
|
| ML classifiers |
Term frequency (TF), Mutual information |
Social media posts |
Logistic regression, Decision trees, XGBoost |
Requires large annotated datasets |
Medhat, Hassan & Korashy (2014), Liu (2022)
|
| Unsupervised learning |
Unsupervised feature extraction |
Customer reviews |
K-means, LDA |
Lexicon-based methods may be limited in domain adaptation |
Cambria et al. (2013), Al-Ghuribi, Noah & Tiun (2020)
|
| Deep learning |
Word embeddings |
News articles |
LSTM, CNN |
High computational cost |
Zhang, Wang & Liu (2018), Liu & Shen (2020)
|
| Transfer learning |
Pre-trained embeddings |
Tweets, Product reviews |
BERT variants |
May require fine-tuning for specific domains |
Tao & Fang (2020), Tan et al. (2022)
|
| Hybrid models |
Ensemble learning |
Tweets, Reviews, Emails |
SVM, LSTM, CNN |
Ensemble models can be computationally expensive |
Dang, Moreno-García & De la Prieta (2021), Janjua et al. (2021)
|
