The Effect of Context on Semantic Similarity Measurement

Similarity measurement is currently being established as a method to explore content on the Semantic Web. Semantically annotated content requires formal concept specifications. Such concepts are dynamic and their semantics can change depending on the current context. The influence of context on similarity measurement is beyond dispute and reflected in recent similarity theories. However, the systematics of this influence has not been investigated so far. Intuitively, the results of similarity measurements should change depending on the impact of the current context. Particularly, such change should converge to 0 with a decreasing impact of the respective contexts. To hold up to this assertion, a quantification of the impact of context on similarity measurements is required. In this paper, we use a combination of the SIM-DL theory, which measures similarity between concepts represented using description logic, and a context model distinguishing between internal and external context to quantify this impact. The behavior of similarity measurements within an ontology specifying geospatial feature types is observed under varying contexts. The results are discussed with respect to the corresponding impact values.