Skip to main content
. 2017 Feb 10;2016:560–569.

Table 3.

Differences in the accuracy of the COS classifier for the semi- and fully supervised cases. Only abbreviations with a performance differential of at least 10 percentage points are shown. Sense counts exclude very rare senses.

abbrev. senses B/A acc. A/A acc. diff.
MOM 2 .425 .998 .573
DT 3 .673 .956 .283
T1 3 .672 .944 .272
PAC 4 .726 .962 .236
PCP 4 .713 .944 .231
AB 3 .760 .958 .198
NAD 2 .745 .942 .197
RT 3 .773 .962 .189
PA 4 .777 .948 .171
DC 4 .755 .904 .146
IT 6 .780 .905 .125
ER 3 .856 .974 .118
GT 3 .825 .940 .115
CD4 3 .865 .972 .107
RA 3 .855 .962 .107
BMP 3 .761 .866 .105