Technologies DIT - University of Trento D ISTRIBUTED O NTOLOGICAL R EASONING : T HEORY , A LGORITHMS , AND A PPLICATIONS

Research on ontologies is becoming a popular topic in various branches of computer science. Their importance has been recognized in knowledge representation, natural language processing, knowledge management, multi-agent systems, intelligent integration of databases, cooperation of distributed applications, and web services. Among others, it gets a special accent and strong push in the context of recently introduced semantic web. During the last years, academia materialized the ontology technology with a variety of theoretical and practical pillars: representational languages, stable underpinning theory, inference algorithms. Developed technology is mature enough to find a wide practical application in a range of scenarios when a single ontology is kept in mind – all the tools are off the shelf. The difficulty and challenge arise when multiple autonomously developed ontologies come into play. This thesis proposes a novel distributed reasoning technique for the setting of multiple ontologies on the semantic web. Unlike standard approaches that are based on integrating several ontologies into a monolithic global ontology and further reasoning in it, the proposed technique adopts a peer-to-peer approach wherein the reasoning results from a coordination of local reasoners associated to each autonomous ontology. Theoretically, the approach is grounded on a Distributed Description Logics framework (DDL). The thesis elaborates inference mechanisms in DDL and introduces sound and complete distributed tableaux algorithms for reasoning with multiple ontologies expressed in SHIQ Description Logic. To estimate the feasibility of the proposed technique and enable its practical deployment on the semantic web, the thesis proposes and implements a distributed reasoning engine entitled DRAGO (Distributed Reasoning Architecture for Galaxy of Ontologies). The system has been evaluated over several application scenarios and demonstrated the suitability and promising results for application of the proposed technique on the semantic web.