Table 7.

Recommendation performance values of the classifiers trained with the MGI data set, based on the first 3 weeks of 2020

		Naïve Bayes classifier		BERT classifier
	n (no. of weekly articles in MGI)	Recall@n	Precision@50	Recall@n	Precision@50
First week of 2020	505	102/505a = 0.20	27/50 = 0.54	170/505 = 0.34	37/50 = 0.74
Second week of 2020	151	52/151 = 0.34	20/50 = 0.40	49/151 = 0.32	30/50 = 0.60
Third week of 2020	184	28/184 = 0.15	19/50 = 0.38	63/184 = 0.34	31/50 = 0.62

^aThe number of weekly Emati recommendations < n.