Table 1.

A general comparison of similar methods in recent years.

References	Method	Language	Dataset	General result
Pathak et al. [47]	Deep learning + topic modeling	English	Facebook, Ethereum, Bitcoin, SemEval-2017	Facebook-0.79, Ethereum-0.844, Bitcoin-0.817, SemEval-2017-0.889
Tang et al.[62]	Topic modeling	English	Amazon, Yelp	Amazon-0.82, Yelp-0.84
Kalarani and Selva Brunda [49]	Joint sentiment-topic features + POS tagging + SVM and ANN	English	Balanced dataset, unbalanced data	SVM-0.84, ANN-0.87
Farkhod et al. [50]	Topic modeling	English	IMDB	IMDB-F1 score-70.0
Fatemi and Safayani [51]	Topic modeling + restricted Boltzmann machine	English	20-Newsgroups (20NG), movie review (MR), multidomain sentiment (MDS)	Perplexity: MR: 406.74
Pathik and Shukla [53]	Deep learning + topic modeling	English	Yelp, Amazon, IMDB	Yelp- 0.75, Amazon-0.76, IMDB- 0.82
Sengupta et al.[54]	Topic modeling	English	Movies, Twitter	Perplexity: Movies- 3834.7, Twitter- 280.75
Huang et al.[56]	Deep learning	English	IMDB, Yelp	IMDB-0.963, Yelp-0.735
Özyurt and Akcayol [57]	Topic modeling	English + Turkish	User reviews in Turkish language about smartphones, SemEval-2016, Task-5 Turkish restaurant reviews	Precision-81.36 Recall-83.43 F-score-82.39
Zhao et al. [58]	Deep learning	English	Amazon review	CNN-87.7, LSTM-87.9
Rao et al. [59]	Deep learning	English	Yelp 2014, 2015, IMDb	Yelp2014–63.9 Yelp2015–63.8, IMDb-44.3
Naseem et al. [60]	Deep learning	English	Airline dataset	Airline dataset = 0.95
Basiri et al. [61]	Attention-based deep learning	English	Sentiment140, Airline, Kindle dataset, movie review	Kindle dataset = 0.93, Airline = 0.92, movie review = 0.90, Sentiment140 = 0.81

The proposed models are deeply described step-by-step in the next section.