|Abstract of Paper|
An Abduction-based Method for Index Relaxation in Taxonomy-based Sources
by Carlo Meghini, Yannis Tzitzikas and Nicolas Spyratos
Indexing accuracy and consistency are difficult to maintain. Imposing a standard indexing language (like the terminology of a taxonomy) tends to improve consistency, but it does not improve accuracy. As a consequence, the extraction of information from a source containing term-classified objects is plagued with uncertainty. Commonly, this uncertainty is dealt with in a quantitative way, by means of numerical methods. In the present paper we deal with uncertainty in a qualitative way. We view an information source as an agent, operating according to an open world philosophy. The agent knows some facts, but is aware that there could be other facts, compatible with the known ones, that might hold as well, although they are not captured for lack of knowledge. These facts are, indeed, possibilities. Here we view possibilities as explanations of knowledge facts and resort to abduction in order to define precisely the possibilities that we want our system to be able to handle. In particular, we introduce an operation that extends a taxonomy-based source with possibilities, and then study the property of this operation from a mathematical point of view.