Mixed Script Identification Using Automated DNN Hyperparameter Optimization

Table 7

Comparison of accuracy.

Reference	Technique	Dataset
		Eng-Roman Urdu	Saraiki-Hindi	Ben-Hindi	Eng-Bengali	Saraiki-English	Saraiki-Roman Urdu	Saraiki-Bengali	Eng-Ben-Sar-Hin-RU
This paper	GloVe-WCF-LSTM	0.561	0.475	0.553	0.593	0.685	0.772	0.562	0.462
	GloVe-WCF-Bi-LSTM	0.864	0.801	0.754	0.812	0.79	0.834	0.831	0.68
	GloVe-WCF-GRU	0.809	0.829	0.845	0.812	0.803	0.89	0.88	0.783
	GloVe-WCF-Bi-GRU	0.882	0.901	0.892	0.899	0.864	0.895	0.861	0.85
[36]	GloVe-LSTM	—	—	86.97	88.27	—	—	—	—
	GloVe-Bi-LSTM	—	—	85.49	87.57	—	—	—	—
[57]	Char-LSTM	—	—	57.8	—	—	—	—	—
	Subword-LSTM	—	—	69.7	—	—	—	—	—