Table 4.

F-measure (%) obtained with each test configuration

	Baseline	#1 filt	#2 gen	#3 d-lex	#4 unk	#5 missp	#6 d-rule	#7 other-lex	#8 other-rule
Horizontal level	77.3	78.0	77.4	75.8	71.5	77.4	78.3	77.6	78.1
Medications	79.8	81.0	80.3	77.1	74.3	80.1	81.7	80.1	80.3
Dosages	80.4	80.4	80.2	79.8	73.3	80.3	80.6	80.3	81.1
Modes	83.6	83.5	83.3	83.2	79.0	83.3	84.0	84.7	85.2
Frequencies	82.7	82.9	82.8	82.1	76.1	82.8	83.3	83.1	84.1
Durations	39.4	40.0	39.1	39.1	34.4	39.4	39.6	39.4	39.5
Reasons	29.9	31.6	29.8	28.6	26.7	29.9	30.1	29.9	30.0

1filt, drug lexicon filtering; #2 gen, generic expressions of drugs; #3 d-lex, additional drug lexicons; #4 unk, unknown drug name recognition; #5 missp, misspellings in drug names; #6 d-rule, drug name extraction rule refinement; #7 other-lex, associated information lexicon extension; #8 other-rule, associated information extraction rule improvement.