ContextAD: Context-Aware Acronym Disambiguation with Siamese BERT Network

Table 6

Overview of the expanded dataset.

	Train	Validation	Test	Total	Growth rate
Original dataset	45,031	5,003	6,189	56,223	—
Raw experimental data	203,438	22,779	28,286	254,503	—
Num = 3	220,136	24,590	30,463	275,189	7.70%
Num = 4	244,863	27,288	33,777	305,928	19.41%
Num = 5	275,234	30,602	37,886	343,722	33.94%
Num = 10	453,200	50, 334	62, 340	565, 874	50.23%