Table 7.

Intrinsic (left 5 columns) and extrinsic scores (right 4 columns) of the biomedical representation models trained using different window sizes

Window Size	UMN-rel(ρ)	UMN-sim(ρ)	MayoSRS (ρ)	Bio-SimVerb(ρ)	Bio-SimLex(ρ)	BC4CHECMD (F-score)	BC2GM (F-score)	AnatEM (F-score)	JNLPBA (F-score)
1	0.5317	0.5759	0.5551	0.4594	0.7294	81.51	70.06	82.16	65.34
2	0.563	0.6144	0.6238	0.4696	0.7207	81.44	70	82.21	65.51
4	0.5768	0.6247	0.581	0.464	0.7188	81.5	70.04	82	65.75
5	0.5767	0.6271	0.5744	0.4638	0.7151	81.37	70.20	81.32	65.16
8	0.582	0.6377	0.5975	0.4611	0.7086	81.24	69.56	80.99	65.53
16	0.5888	0.6431	0.6123	0.4667	0.7034	81.02	69.39	80.72	64.78
20	0.5896	0.6418	0.6319	0.4584	0.7031	81.12	69.62	80.49	65.19
25	0.6018	0.6489	0.6188	0.4519	0.7004	81.07	69.93	80.92	65.14
30	0.6007	0.6457	0.6486	0.4502	0.7043	80.71	69.2	81.03	64.79

The bolded text implies the best performing models of their kind