An annotation and modeling schema for prescription regimens

Table 7 Accuracy results for tag attributes

Choice Attributes	Baseline	Literal Fallback	Levenshtein Fallback	Classifier
form	0.597	0.944	0.965	0.944
route	0.339	0.667	0.738	0.887
side	0.792	0.935	0.173	0.935
unit	0.477	0.831	0.869	0.823
Attributes of FREQ	Baseline	Hybrid	Classifier
every	0.808	0.984	0.981
to_every	0.954	1.000	0.986
times_per	0.570	0.927	0.900
to_times_per	0.876	0.949	0.954
unit	0.722	0.976	0.976
Attributes of TIMING	Baseline	Hybrid	Classifier
direction	0.396	0.990	0.990
event	0.406	0.919	0.919
offset	0.480	0.883	0.742
to_offset	0.866	0.950	0.950
offset_unit	0.430	0.826	0.826
Numeric Attributes	Un-normalized	Normalized
amt	0.160	0.927
to_amt	0.882	0.857
num	0.115	0.885
to_num	0.836	0.885
quantity	0.889	1.000
to_quantity	1.000	1.000
Boolean Attributes	Baseline	Classifier
refill	0.333	1.000
subst	0.500	0.615

ISSN: 2041-1480