ContextAD: Context-Aware Acronym Disambiguation with Siamese BERT Network

Table 4

Overview of the expanded validation and test dataset.

	Train	Validation	Test	Total	Growth rate of test dataset
Original dataset	45,031	5,003	6,189	11,192	—
Raw experimental data	203,438	22,779	28,286	254,503	—
Num = 3	203,438	24,590	30,463	258,491	7.70%
Num = 4	203,438	27,288	33,777	264,503	19.41%
Num = 5	203,438	30,602	37,886	271,926	33.94%
Num = 6	203,438	34,303	42,493	280,234	50.23%
Num = 7	203,438	38,111	47,231	288,780	66.98%
Num = 8	203,438	42,251	52,347	298,036	85.06%
Num = 9	203,438	46,582	57,671	307,691	103.89%
Num = 10	203,438	50,334	62,340	316,112	120.39%