Journal of Big Data

Table 7 Results of BLSTM using different pre-trained word embeddings on the first and second data sets

From: Investigating the impact of pre-processing techniques and pre-trained word embeddings in detecting Arabic health information on social media

	First data set				Second data set
	Precision	Recall	F₁ score	Accuracy	Precision	Recall	F₁ score	Accuracy
AraVec Skip-Gram	89.14	88.85	89	93.3	82.19	63.49	71.64	90.5
AraVec CBOW	87.09	86.23	86.66	91.9	79.35	65.08	71.51	90.27
Mazajak Skip-Gram	90.27	88.2	89.22	93.5	75.81	74.6	75.2	90.7
Mazajak CBOW	90.9	88.52	89.7	93.8	88.19	59.26	70.89	90.8
fastText	89	87.54	88.26	92.9	83.2	57.67	68.13	89.8
ArWordVec Skip-Gram	89.6	87.54	88.56	93.1	71.76	64.55	67.97	88.5
ArWordVec CBOW	91.87	85.25	88.44	93.2	76.92	68.78	72.63	90.2

Bold numbers indicate the best value while underlined numbers represent the second-best value

Back to article page