Eugeniu Arbuleac

Ontology negotiation

Goals, Requirements and Implementation

! ! ! ! !

Introduction! Conceptual framework! Operational framework! Communication ! Conclusion


! !

MASs - Multi agent systems ! Communication in heterogenous MASs is difficult due to lack of shared ontologies! Ontology negotiation = an integrated approach that allows agents to build semantically integrated systems by sharing parts of their ontologies


=> Important goals and requirements for any ontology negotiation protocol to work .Introduction ! Ontology negotiation involves a combination of agent communication protocol and ontology alignment protocol.

Agents share a common ontology! MASs .Introduction . all-purpose ontologies => disadvantageous for problem solving skills of an agent (his own understanding of the world) ! .difficult to apply normal communication protocols! Current solutions: Merged ontologies.Problems ! ! Assumption .

large scale problems => Ontology agent should compute mappings between every ontology on runtime .time/cost consuming ! ! ! .Introduction ! Ontologies alignment = set of mappings that allows agents to use different ontologies and keep their individual ontologies ! Disadvantages: Pre-built set of mappings! Solved: Ontology agents ! ! Not compatible with real world.

agents solve their communication problems at runtime by exchanging parts of their ontologies! Agents should be able to automatically match ontologies or detect if their ontologies are insufficiently aligned for successful communication! Ontology negotiation = agent communication protocol. automatic ontology matching and ontology mismatches detection ! ! .Introduction ! Ontology Negotiation .

Ontology negotiation ! uses shared ontology ! = communication ! vocabulary (cv)! shared ontology! = “agreement”! distributed nature ! ! .

Efficiency! .Simplicity ! ! .Ontology negotiation ! Cooperation between agents with different interests through negotiation protocols! (ex. buyers & sellers decide upon price)! Characteristics:! .

Ontology negotiation ! Efficiency = minimal cv construction. acceptable solution = sound and lossless communication sound = quality of information exchange! lossless = quantity of information exchange ! ! ! .

Ontology negotiation ! Simplicity = low computational and bandwidth demands on agents! Concept learning . by solving each communication problem through sharing parts of ontology ! ! .computational expensive => learn on as-need basis = lazy ontology alignment! Lazy ontology alignment helps incrementally build a solution for semantic integration problem between agents.

Conceptual Framework .

Conceptual framework ! two agents a1 and a2! ! ! two Ontologies O1 and O2! Ox—Oy => Oy is included in Ox .

Ontology ! Conceptualisation = meanings that are used by the agent to represent knowledge! Specifies the conceptualisation with a set of symbols C and a reflexive transient function ! .

Local ontology ! Local ontology = a vocabulary which the agent uses to reason with assertion knowledge! Stores agent’s assertion knowledge efficient and useful! Assertion knowledge base A = SUM ( membership statements representing the belonging of one individual to a concept .disjoint sets C1 and C2 for concepts ) ! ! . without conflicting concepts .

God’s eye view ontology ! The ontology O that arise if the local ontologies of the agents were combined! Virtual ontology! Any other ontology in this system is included in O ! ! .

where O1-cv U O2-cv = Ocv .Ontologies for alignment ! Ontology Ocv align the agents’ local ontologies mapping their local ontologies to communication vocabulary ! O1-cv and O2-cv.

someone in the group knows it.Knowledge ! Local knowledge . or knowledge is distributed within the group ! ! .only by one agent! Common knowledge every agent! Implicit group knowledge .

change over time ! .Dynamics ! Static ontology .do not change over time! Dynamic ontology .

changes => O1-cv and O2-cv changes! O1. O2 and O1-2 remain still as static ontologies ! .Concept learning ! Ocv .

The receiver does the opposite process to understand the sender’s message .Communication ! The sender makes itself understandable to the receiver by translating the message stated in terms of its local ontology to a message stated in terms of communication vocabulary.

it follows that a is member of superconcept of c . the interpretation of the message by the hearer must follow from what speaker intended to convey! ! When a is member of a concept c.Requirements for normal communication ! Soundness = quality of communication.

Requirements for normal communication Lossless communication = no information is lost in the process of translating the message to and from communication vocabulary! ! .

Operational framework .

Operational framework ! ! Description logic to implement ontologies! Meets the requirements regarding knowledge distribution! More powerful than Conceptual Framework ! .

! TBox = a set of terminological axioms=agent’s ontology! ABox = a set of membership statements that specify which of the individuals belong to which concepts ! ! .A).Description logic ! Description logic knowledge base = a tuple (T. containing TBox and ABox.

Syntax ! Used to implement a set of concept names C.Language .atomic concept. L(Ca) . Ca .

Semantics .

Local and common KB ! Local knowledge of an agent over and ontology can be established using TBoxes! Common knowledge is established using another TBox! => TBoxes fully implement the agent’s knowledge over ontologies ! ! .

Implicit group knowledge ! Agents no more about their local ontologies => using the action Classify! This way the union of TBoxes will give the Implicit group knowledge of the full ontology O1-2 ! .

Communication .

receive actions are used for message transmission and interception ! .Communication ! Actions that describe the communicative abilities of the agents! Send.

Implementation of concept learning .

Ontology negotiation protocols Protocol 1 .

Ontology negotiation protocols Protocol 2 .

Ontology negotiation protocols Protocol 3 .

Ontology negotiation protocols Protocol 3 .

Ontology negotiation protocols Evaluation of the protocols .

Conclusion .

on as-need basis! Small communication vocabulary => easy to learn and to process ! ! .Conclusion ! An ontology negotiation protocol should enable soundness and lossless communication between agents! Lazy build of solution.