Proceedings Abstracts of the Twenty-Third International Joint Conference on Artificial Intelligence

Syntactic Labelled Tableaux for Łukasiewicz Fuzzy ALC / 962
Agnieszka Kułacka, Dirk Pattinson, Lutz Schröder

Fuzzy description logics (DLs) serve as a tool to handle vagueness in real-world knowledge. There is particular interest in logics implementing Lukasiewicz semantics, which has a number of favourable properties. Current decision procedures for Lukasiewicz fuzzy DLs work by reduction to exponentially large mixed integer programming problems. Here, we present a decision method that stays closer to logical syntax, a labelled tableau algorithm for Lukasiewicz Fuzzy ALC that calls only on (pure) linear programming, and this only to decide atomic clashes. The algorithm realizes the best known complexity bound, NEXPTIME. Our language features a novel style of fuzzy ABoxes that work with comparisons of truth degrees rather than explicit numerical bounds.