Table 2.

Error analysis evaluation results on different types of error occurrence on the test dataset.

Sl. No.	Sources of error	True value in data	Observed value in data	Error percentage
1.	Entity detection error	633,074*	582,428	8.00%
2.	Entity absent in text	633,074*	615,650	2.75%
3.	Failure to detect entity	633,074*	609,413	3.73%
4.	Entity normalisation error
a.	Gene normalization error	42,607	50,336	18.14%
b.	Disease normalization error	71,704	92,481	28.97%
c.	Drug normalization error	11,033	14,563	31.99%

$\text{Error percentage}=\frac{\left|\text{Observed value-True value}\right|}{\text{True value}}\times 100(1)$

PharmGKB has been considered as the gold standard dataset for all the comparisons. *in total PGx corpus extracted from MEDLINE. The error percentage has been calculated according to the formula 1.