Table 3.

Top reported system performance for studies predicting the gender of Twitter users using deep learning machine learning (ML) methods. Result metrics are reflected in this table as reported in the original publications and are not comparable to each other.

Study	Language	ML method	Reported performance
			F1-score	Accuracy
Ardehaly and Culotta [52], 2017	English	Deep LLPa	0.96	N/Ab
Geng et al [71], 2017	English	Ensemble: LDAc and CNNd	N/A	0.87
Kim et al [83], 2017	English	GRNNe	N/A	0.68
Vijayaraghavan et al [108], 2017	English	DMTf	0.89	N/A
Wang et al [111], 2017	N/A	CNN	0.91	0.9
Bayot and Goncalves [55], 2017	English and Spanish	CNN	N/A	0.59-0.72
Bsir and Zrigui [57], 2018	Arabic	GRUg	N/A	0.79
Wood-Doughty et al [115], 2018	English	RNNh	0.84	0.84
Bsir and Zrigui [58], 2019	Arabic	BILSTMi with attention	N/A	0.82
Hashempour [75], 2019	Portuguese, French, Dutch, Spanish, German, and Italian	FFNNj	N/A	0.84-0.86
Wang et al [112], 2019	Multilingual	mmDNNk	0.92	N/A
ElSayed and Farouk [68], 2020	Egyptian and Arabic dialects	Multichannel CNN-biGRUl	N/A	0.84-0.91
Imuede et al [93], 2020	English	DNNm	N/A	0.68
Zhao et al [121], 2020	English	CNN	0.80	N/A
Yang et al [117], 2021	English	Ensemble: M3n and SVMo	0.95	0.94

^aLLP: learning with label proportions.

^bN/A: not applicable.

^cLDA: latent Dirichlet allocation.

^dCNN: convolutional neural network.

^eGRNN: graph recurrent neural network.

^fDMT: deep multimodal multitask.

^gGRU: gated recurrent network.

^hRNN: recurrent neural network.

ⁱBILSTM: bidirectional long-term short-term memory.

^jFFNN: feed forward neural network.

^kmmDNN: multimodal deep neural network.

^lbiGRU: bidirectional gated recurrent unit.

^mDNN: deep neural network.

ⁿM3: multimodal, multilingual, and multi-attribute system.

^oSVM: support vector machine.