TABLE V:

Ablation study - Bold entries indicate biggest drop in MAP and blue entries correspond to best MAP values

Models	Optimized ARS weights					MAP
	SDM score	QAMat	MeSH Dist.	SemMedDB1	SemMedDB2	batch1	batch2	batch3	batch4	batch5
All	0.3043	0.4747	0.0846	0.0102	0.1264	0.4438	0.4780	0.4534	0.4388	0.3722
– QAMat	0.4515	-	0.2848	0.0718	0.1919	0.4202	0.4721	0.4227	0.4145	0.3604
– MeSH Dist.	0.3279	0.5771	-	0.0309	0.0642	0.4410	0.4659	0.4476	0.4307	0.3614
– SemMedDB	0.2896	0.5365	0.1739	-	-	0.4352	0.4680	0.4329	0.4161	0.3521
Baseline	1.0000	-	-	-	-	0.4279	0.4709	0.4306	0.4219	0.3505
(a) Results on year 2016 datasets
Models	Optimized ARS weights					MAP
	SDM score	QAMat	MeSH Dist.	SemMedDB1	SemMedDB2	batch1	batch2	batch3	batch4	batch5
All	0.1665	0.7298	0.0411	0.0062	0.0564	0.4075	0.4363	0.4534	0.3891	0.2316
– QAMat	0.5243	-	0.1832	0.2406	0.0520	0.3782	0.4190	0.4372	0.3690	0.2181
– MeSH Dist.	0.3121	0.5494	-	0.0638	0.0747	0.3956	0.4221	0.4471	0.3772	0.2144
– SemMedDB	0.2970	0.5220	0.1811	-	-	0.3808	0.4316	0.4459	0.3758	0.2154
Baseline	1.0000	-	-	-	-	0.3959	0.4176	0.4378	0.3746	0.2133
(b) Results on year 2017 datasets