. Author manuscript; available in PMC: 2009 Jan 1.

Published in final edited form as: Artif Intell Med. 2007 Nov 28;42(1):13–35. doi: 10.1016/j.artmed.2007.10.001

Table 18.

Comparison of features for challenge corpus. For all pairs of features, the differences between F-measures for PHI and the differences between F-measures for non-PHI are significant at α = 0.05. Best F-measures are in bold.

Feature	Class	Precision	Recall	F-measure
Target words	Non-PHI	96.90%	99.87%	98.36%
Target words	PHI	96.05%	49.56%	65.38%
Lexical bigrams	Non-PHI	97.34%	99.69%	98.50%
Lexical bigrams	PHI	91.99%	56.87%	70.29%
Syntactic bigrams	Non-PHI	97.50%	99.74%	98.61%
Syntactic bigrams	PHI	93.44%	59.61%	72.79%
POS information	Non-PHI	96.04%	99.42%	97.70%
POS information	PHI	79.33%	35.24%	48.80%
Dictionary	Non-PHI	94.26%	99.90%	96.99%
Dictionary	PHI	69.70%	3.79%	7.19%
MeSH	Non-PHI	94.05%	100%	96.93%
MeSH	PHI	0%	0%	0%
Orthographic	Non-PHI	96.05%	99.60%	97.79%
Orthographic	PHI	84.67%	35.30%	49.83%