Semantic Measures

Similarity and Semantic Relatedness

This section discusses the three different types of semantic measures. Depending on the applications and the developers’ needs, proposals may be semantic relatedness measures, semantic similarity measures or semantic distance. The semantic measure, which allows to quantify the distance between the meanings of two concepts, is a generic term covering several concepts (Budanitsky & Hirst, 2006) (Gracia & Mena, 2008):

• •

  Semantic relatedness covers all possible semantic relationships. This is a broader measure than the measure of semantic similarity. Indeed, the terms which do not share a common meaning can be considered semantically close, as they can be linked by a meronym or antonym relationship. They can also be linked by a functional relationship or frequent association relationship (e.g. “car” and “road,” “lion” and “Africa”).

• •

  Semantic similarity is a special case of semantic relatedness. This distance uses only synonymy, hyponymy and hyperonymy relationships to determine whether two words share common characteristics.

• •

  Semantic distance is often viewed as the inverse of semantic relatedness or semantic similarity. If the proximity increases, the semantic distance decreases. In most cases, the two “visions” of the term “distance” are compatible. However, there are exceptions. For example, antonym concepts are semantically dissimilar but still very close, due to the antonymous relationship. Generally, it is accepted that semantic distance is the inverse of semantic relatedness.