Abstractive Arabic Text Summarization Based on Deep Learning

Table 3

The result of applying GRU, LSTM, and BiLSTM with different layers at the encoder to the AHS dataset.

Model	No. of layers	ROUGE-1			ROUGE-2			ROUGE-L			BLEU
		Precision	Recall	F1	Precision	Recall	F1	Precision	Recall	F1
GRU	1	49.44	44.12	45.17	8.0	7.13	7.35	34.05	31.31	29.81	0.32
	2	50.35	45.65	46.44	9.1	8.16	8.4	34.52	32.2	30.61	0.34
	3	49.87	44.0	45.37	8.08	7.24	7.47	34.53	31.4	30.14	0.32
LSTM	1	53.76	48.87	50.1	11.78	10.73	10.99	37.1	34.1	33.44	0.38
	2	54.82	49.53	50.93	12.27	11.29	11.51	37.71	34.46	33.93	0.39
	3	53.04	47.99	49.3	11.47	10.29	10.64	37.15	33.78	33.26	0.37
BiLSTM	1	54.16	49.84	50.8	12.65	11.63	11.89	37.51	34.87	34.05	0.39
	2	55.12	50.2	51.46	13.13	11.92	12.25	38.0	35.04	34.44	0.40
	3	54.95	50.48	51.49	13.1	12.01	12.27	37.84	35.19	34.37	0.41

The highest result is given in bold.