Table 1 Summary of key deep learning-based sentiment analysis studies: methods, datasets, and reported performance.
Author(s) | Approach/model | Dataset/domain | Reported performance | References |
|---|---|---|---|---|
Wang et al. | CNN-LSTM hybrid | Yelp Reviews | Improved classification accuracy | |
Smaid et al. | BiLSTM-CRF + AB-LSTM-PC | Arabic Hotel Reviews | Enhanced aspect & sentiment detection | |
Jelodar et al. | LSTM + NLP | COVID-19 Tweets | 81.15% accuracy | |
Basiri et al. | ABCDM (CNN + GRU + Attention) | Tweets and Product Reviews | Outperformed baselines | |
Jianqiang & Xiaolin | CNN with Word Embeddings | Accuracy improvement over baseline | ||
Xu et al. | BiLSTM + Weighted Word Vectors | User-generated Comments | Enhanced sentiment representation | |
Muhammad et al. | LSTM + Word2Vec | Hotel Reviews | 85.96% accuracy | |
Haque et al. | CLSTM (CNN + LSTM) | Bengali Social Media Comments | 85.8% accuracy, F1: 0.86 | |
Deng et al. | SSALSTM (Self-Attention LSTM) | E-commerce Dataset | Improved class balance handling | |
Afidah et al. | LSTM + CNN + Word2Vec | Indonesian Tourist Reviews | 97.17% accuracy |
