An Efficient Algorithm for Computing Elected Assertions in Partially Preordered Ontologies
Sihem Belabbès, Salem Benferhat · 2021 IEEE 33rd International Conference on Tools with Artificial Intelligence (ICTAI) · 2021
Handling inconsistency in formal ontologies is crucial for facilitating meaningful query answering. Arguably the most popular approach for resolving inconsistency amounts to repairing the dataset in terms of the semantic knowledge encoded in the ontology. This has given rise to many inconsistency- tolerant semantics, like the well-known IAR (Intersection of ABox Repair) semantics, which produces a single consistent subset of the dataset, and that can be queried. Several frameworks additionally consider a preference relation over the data pieces (called assertions), such as the Elect method, which generalizes the IAR semantics to capture a partial preorder. Elect also computes a single consistent subset of the dataset using the notion of elected assertions. Basically, an assertion is elected if it is strictly preferred to all the assertions that conflict with it. However, Elect requires the prior computation of all the conflicts between the assertions. In this paper, we propose a new algorithm for computing the set of elected assertions, without exhibiting all the conflicts. Our algorithm is based on a new characterization of the set of elected assertions using the IAR semantics.