Table 2.

Results compared with the existing models.

Model	Accuracy (%)	Precision (%)	Recall (%)	F1 (%)
LIWCa,b	70	74	71	72
LDAb,c	75	75	72	74
Unigramb	70	71	95	81
Bigramb	79	80	76	78
LIWC + LDA + unigramb	78	84	79	81
LIWC + LDA + bigramb	91	90	92	91
LSTMd	87.03	90.30	91.67	90.98
Bi-LSTMe	86.46	88.08	95	91.41
Bi-LSTM + Attf	88.59	90.41	94.96	92.63
EANg (our model)	91.3	91.91	96.15	93.98

^aLIWC: Linguistic Inquiry and Word Count.

^bIndicates that the results are shown in the literature [4].

^cLDA: latent Dirichlet allocation.

^dLSTM: long short-term memory.

^eBi-LSTM: bidirectional long short-term memory.

^fAtt: attention mechanism.

^gEAN: emotion-based attention network.