Semantic similarity-based approach to enhance supervised classification learning accuracy

This brief communication discusses the usefulness of semantic similarity measures for the evaluation and amelioration of the accuracy of supervised classification learning. It proposes a semantic similarity-based method to enhance the choice of adequate labels for the classification algorithm as well as two metrics (SS-Score and TD-Score) and a curve (SA-Curve) that can be coupled to statistical evaluation measures of supervised classification learning to take into consideration the impact of the semantic aspect of the labels on the classification accuracy.