A Parametric Similarity Method: Comparative Experiments based on Semantically Annotated Large Datasets