Table 3.

Macro-averaged performance comparing the pLTR method against several baseline approaches, with the highest value in bold

Setup	Precision	Recall	F-measure
NLM Lexical Normalization	0.213	0.718	0.316
MetaMap	0.510	0.702	0.559
Inference method	0.597	0.731	0.637
BANNER + Lucene	0.662	0.714	0.673
BANNER + cosine similarity	0.692	0.732	0.711
DNorm (BANNER + pLTR)	0.828	0.819	0.809