Refining the Notions of Depth and Density in WordNet-based Semantic Similarity Measures