Mixed Script Identification Using Automated DNN Hyperparameter Optimization

Table 6

Accuracy comparison with parameter optimization though random estimator.

Glove embedding with random estimator (random suggest)
Dataset	Classification algorithms
	LSTM	Bi-LSTM	GRU	Bi-GRU
Eng-Roman Urdu	0.561	0.864	0.809	0.882
Saraiki-Hindi	0.475	0.801	0.829	0.901
Bengali-Hindi	0.553	0.754	0.845	0.892
Eng-Bengali	0.593	0.812	0.812	0.899
Saraiki-English	0.685	0.79	0.803	0.864
Saraiki-Roman Urdu	0.772	0.834	0.89	0.895
Saraiki-Bengali	0.562	0.831	0.88	0.861
Eng-Bengali-Saraiki-Hindi-Roman Urdu	0.462	0.68	0.783	0.85

Bold indicates the maximum performance of the algorithm.