Αναπαράσταση Γνώσης και Συλλογιστική

Επιστημονικός υπεύθυνος: Καθ. Δημήτρης Πλεξουσάκης

Objectives

Most corporate and scientific worlds are by no means simple and the knowledge that must be represented and managed about them is fairly complex and dynamic in nature. Not only adequate representation formalisms are required, but also elaborate reasoning techniques for producing additional knowledge. The ISL has been pursuing and plans to continue pursuing basic research in knowledge representation and reasoning. Particular objectives are to:

  • Develop rule-based reasoning languages and systems for the semantic web
  • Integrate methods from nonmonotonic reasoning with semantic web languages (RDF/S and/or OWL)
  • Study the revision of description logic based knowledge bases, and apply the results and techniques to ontology revision and the integration of inconsistent ontologies
  • Apply the above methods to e-Commerce, pervasive computing, and IT security.

Motivation

The semantic web is a major international effort aiming at extending the current state of the World Wide Web though the use of semantics. It is steadily gaining momentum both in academia and industry, and has already developed a number of web standards, most notably RDF/S and OWL. The current web ontology language is based on description logics. Now that this work is essentially completed, the focus is turning towards the use of rules, to allow for increased expressive power and to utilize existing rule technology for systems development. Work in this direction mainly studies the integration of rules and description logics (e.g. SWRL), which poses many computational and complexity problems.

Our work in this direction will particularly study rule systems that can deal with incomplete and inconsistent information, using methods from nonmonotonic reasoning. Incompleteness is an important issue in Web systems, since information may be missing but decisions still have to be made (e.g. by automated agents). One particular focus of our work is the use of computationally simple, rule-based nonmonotonic systems that allow for good performance while still being able to provide advanced reasoning capabilities. This work will initially focus on extending RDF Schema with monotonic and nonmonotonic rules.

The semantic web initiative, but increasingly also information systems, relies on ontologies as explicit formalizations of conceptualizations. In general, different systems may commit to different ontologies, but may have to work together. Putting together ontologies and resolving inconsistencies is a major challenge. We intend to attack this problem through the use of methods from belief revision, in particular with AGM belief revision that is the best known and most broadly used approach.

Apart from conducting theoretical work and developing prototypes, we intend to apply the above techniques in real world problems. One prime candidate is e-commerce, where we intend to apply semantic web and rules technology to the problems of brokering, automated agent negotiation, auctioning, and electronic contracts. The main motivation is to support (semi-)automation of these processes through the use of intelligent software agents, the behavior of which is governed by declarative policies. A main advantage of having declarative descriptions of agent behavior is the ability of the agents to explain their behavior, which has the potential to increase users' trust in them.

Another promising application is pervasive computing, which is considered to be one of the key issues in the further development and use of information and communication technologies. Any pervasive application must be context-aware. The most modern approaches use ontology-based models of context and aim to better support interoperation by formally defining common concepts and the relations and mapping between them. Once context information has been modeled, it has to be processed. Such reasoning goes well beyond the capabilities of reasoning methods associated with RDF Schema and OWL. In our work we wish to apply our rule-based systems to context modeling, taking advantage of the low computational complexity of the methods we are developing and their capability to reason with incomplete and inconsistent information. Such information often occurs in pervasive environments, e.g. when sensor data are inconclusive or contradictory.

Another application we wish to explore is IT security, an aspect that is becoming increasingly important in the use of information and communication technologies. Our aim is to study the use of logical languages and systems for the expression, verification, and application of declarative security policies.

Accomplishments

  • A defeasible reasoning system for reasoning about monotonic and nonmo0notonic rules and priorities among them
  • First prototypes of the above systems, incorporating RDF/S knowledge
  • A study of applying belief revision theory to description logics, at present focusing on AGM compliance of description logics
  • A simple agent-based prototype for semantic brokering
Όροι Χρήσης | Πολιτική Προστασίας Δεδομένων Προσωπικού Χαρακτήρα