Using Ontological Contexts to Assess the Relevance of Statements in Ontology Evolution

1 minute read

by Fouad Zablith, Mathieu d'Aquin, Marta Sabou, and Enrico Motta

This work describes a method for judging the relevance statements suggested by ontology evolution tools for an existing ontology.

The relevance assessment process consists of the following steps:

  1. locate possible online ontologies for contextualization based on the statements subject and object (=involved concepts).
  2. identify the context C of the new statement based on online ontologies using the Jaro-Winkler string similarity metric (extract all nodes contained both ontologies).
  3. estimate the relevance of the statement using (a) overlap analysis, or (b) a pattern-based relevance assessment.
Overlap Analysis

The overlap analysis computes the relevance based on the overlap between the context and the evolving ontology.

\[conf_{overlap} (c, C, O) = \frac{ |e(C) \hut e(O_t)|}{e(O_t)} \]

A drawback of this method is, that it considers all statements equally important.

Pattern-Based Relevance Assessment

This method considers the structure in which the concepts appear. The authors have identified five patterns and determined the patterns' relative importance for detecting relevant statements.

The evaluation has been performed based on three different domains by

  • automatically determining statements for a given base ontology
  • having domain experts evaluate these statements and computing a per statement relevance score.
  • comparing this score to the suggestion by the metrics introduced above.
The method is highly relevant, but still only deals with quite similar ontologies, which are extended based on a common conceptualization.