PRUV ON THURSDAY, JULY 24TH

ABSTRACT. The use of preferences in query answering, both in traditional databases and in ontology-based data access, has recently received much attention, due to its many real-world applications. In this paper, we tackle the problem of query answering in Datalog+/– ontologies subject to the querying user’s preferences and a collection of subjective reports (i.e., scores for a list of features) of other users, who have their own preferences as well. All these pieces of information are combined to rank the query results. We first focus on the problem of ranking atoms in a database by leveraging reports and customizing their content according to the user’s preferences. Then, we extend this approach to deal with ontological query answering using provenance information. Though the general problem is shown to have an exponential-time data complexity upper bound, we propose a special case that has polynomial time data complexity.

ABSTRACT. Dealing with uncertainty is a very important issue in description logics (DLs) because these latters are designed only for certain knowledge. Therefore, several approaches proposed probabilistic extensions of DLs but most of them are difficult to use and don't support the belief in terminological axioms or use graphical models such as Bayesian Network or based on classical models. In this paper, we present PrDL-lite N bool a new probabilistic extension of DL-lite N bool by supporting the belief interval in a single axiom or a set of axioms connected with conjunction (by ∧) or disjunction (by ∨) operators. The PrDL-lite N bool semantics is based on DL-lite N bool features which are a new alternative semantics for DL-lite N bool having a finite structure and the number of them is always finite unlike classical models. PrDL-lite N bool supports terminological and assertional probabilistic knowledge and the reasoning tasks (satisfiability, deciding the entailment, computing the tight belief interval for entailment) are achieved by solving linear programs. A prototype of the approach is implemented using OWLAPI for knowledge base creation, Pellet for reasoning and LpSolve for solving the linear programs.

ABSTRACT. In a relational setting, the maximum entropy model of a set of probabilistic conditionals can be defined referring to the full set of ground instances of the conditionals. The logic FO-PCL uses the notion of parametric uniformity to ensure that the full grounding of the conditionals can be avoided, thereby greatly simplifying the maximum entropy model computation. In this paper, we describe a system that realises an approach transforming an FO-PCL knowledge base consisting of relational probabilistic conditionals into a knowledge base having the same maximum entropy model that is parametrically uniform. The implemented system provides different execution and evaluation modes, including the generation of all possible solutions, and is available within an integrated development environment for relational probabilistic logic.