Table 3.

Table comparing the performance of different methods based on SemEval-2016 dataset with in-sample, out-of-sample and 10-fold cross-validation accuracy.

	Out-of-sample	In-sample	Cross-validation
	Accuracy	Accuracy	Accuracy	St. dev.
mOnt	78.31%	75.31%	75.31%	0.0144
wOnt	72.80%	70.80%	70.90%	0.0504
sOnt	76.46%	73.92%	73.87%	0.0141
mOnt + CABASC	85.11%	82.73%	80.79%	0.0226
mOnt + LCR-Rot-hop	86.80%	88.21%	82.88%	0.0224
sOnt + CABASC	83.16%	79.53%	72.04%	0.1047
sOnt + LCR-Rot-hop	84.49%	86.07%	79.73%	0.0348

wOnt stands for a similar semi-automatic ontology built with the same methods as sOnt but with words (instead of synsets) as terms.