Table 1. Statistics of the three types of ASC methods.

Model represents the type of deep learning methods adopted by the corresponding published paper. Aspect and position with a check mark (√) denotes the model considering the aspect information and position information, respectively. Attention without X means the model using the attention, CA, GA and DPA indicate “Contact Attention”, “General Attention” and “Dot-Product Attention” respectively. Class-level indicate different modeling paradigms, SeqClass, Seq2Seq, DepenTree and TokenClass indicate “Sequence-level Classification”, “Sequence-to-Sequence”,“Dependency-Tree” and “Token-level Classification”, respectively.

	Model	Attention	Aspect	Position	Modeling paradigms
Based RNN	TNET (Li et al., 2018)	CA	√	√	SeqClass
	RAM (Chen et al., 2017), PBAN (Gu et al., 2018)	GA	√	√	Seq2Seq
	IAN (Ma et al., 2017)	GA	√	X	SeqClass
	ATAE-LSTM (Wang, Huang & Zhao, 2016), DyMemNN (Tay, Tuan & Hui, 2017)	CA	√	X	Seq2Seq
Based GCN	KumaGCN (Chen, Teng & Zhang, 2020)	CA	√	X	DepenTree
	DualGCN (Li et al., 2021)	GA	√	√	DepenTree
	ASGCN (Zhang, Li & Song, 2019b)	DPA	√	√	DepenTree
	BiGCN (Bian et al., 2020), TGCN (Tian, Chen & Song, 2021)	CA	√	√	DepenTree
Based PLMs	Post-Training BERT (Xu et al., 2019)	DPA	X	√	TokenClass
	CG-BERT (Qiu et al., 2020)	DPA	√	√	TokenClass
	BERT-Pair (Wu et al., 2020)	DPA	X	√	TokenClass
	RoBERTa (Dai et al., 2021)	DPA	X	√	TokenClass