ArThUR: A tool for markov logic network

Abstract

Logical approaches-and ontologies in particular-offer a welladapted framework for representing knowledge present on the Semantic Web (SW). These ontologies are formulated in Web Ontology Language (OWL2), which are based on expressive Description Logics (DL). DL are a subset of First-Order Logic (FOL) that provides decidable reasoning. Based on DL, it is possible to rely on inference mechanisms to obtain new knowledge from axioms, rules and facts specified in the ontologies. However, these classical inference mechanisms do not deal with : uncertainty probabilities. Several works recently targeted those issues (i.e. Pronto, PR-OWL, BayesOWL, etc.), but none of them combines OWL2 with Markov Logic Networks (MLN) formalism. Several open source software packages for MLN are available (e.g. Alchemy, Tuffy, RockIt, etc.). In this paper, we present ArThUR, a Java framework for reasoning with probabilistic information in the SW. ArThUR incorporate three open source software packages for MLN, which is able to reason with uncertainty information, showing that it can be used in several real-world domains. We also show several experiments of our tool with different ontologies.