Post on 30-Dec-2015
Non-holistic AgentsNon-holistic Agents
A project ideaA project idea
Patrick De CausmaeckerPatrick De Causmaecker
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
2
Who we areWho we are
Research group within Kaho St-Lieven in Research group within Kaho St-Lieven in Ghent, BelgiumGhent, Belgium
Four years of technology transfer research Four years of technology transfer research in agents technolgyin agents technolgy
Sponsored by the Flemish governmentSponsored by the Flemish government
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
3
AgentsAgents
Agents are specialised in one problem Agents are specialised in one problem domaindomain
They are not designed to understand the They are not designed to understand the whole business model of an application whole business model of an application they are visitingthey are visiting
They have a thorough understanding of They have a thorough understanding of their own field, and bear a model of this their own field, and bear a model of this field (ontology)field (ontology)
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
4
Holistic systemsHolistic systems
The applications they are visiting may be The applications they are visiting may be holistic or notholistic or not
In general, they have to communicate with In general, they have to communicate with a set of applications, more or less a set of applications, more or less integrated, which do solve the actual integrated, which do solve the actual automation problem of the businessautomation problem of the business
In this sense they visit a system bearing a In this sense they visit a system bearing a model of the businessmodel of the business
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
5
Agents on foreign territoryAgents on foreign territory
Agents are designed to serveAgents are designed to serve They must apply their specialised They must apply their specialised
knowledge to boost the performance of knowledge to boost the performance of their client systemstheir client systems
The model of their specialisation will in The model of their specialisation will in general not fit into the application set of the general not fit into the application set of the clientclient
They have to be able to find the crucial They have to be able to find the crucial hooks in the client applicationhooks in the client application
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
6
Goal of the proposalGoal of the proposal
For this they need a mapping methodologyFor this they need a mapping methodology One side of the mapping is the system of One side of the mapping is the system of
the clientthe client The other side is the domain model the The other side is the domain model the
agent is carryingagent is carrying This model may be a (thin) ontology This model may be a (thin) ontology
enabling the agent to communicate with enabling the agent to communicate with other agents from the same domainother agents from the same domain
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
7
Goal continuedGoal continued
The client system will look as a chaotic set The client system will look as a chaotic set of data to the agentof data to the agent
The mapping will require other knowledge The mapping will require other knowledge and expertise about the business domain and expertise about the business domain than the agent knows aboutthan the agent knows about
The agent may have to build on previous The agent may have to build on previous experiences and will be a learning machineexperiences and will be a learning machine
It will perform an analysis of the system in It will perform an analysis of the system in which it arriveswhich it arrives
23-11-00 Semantic Web Technologies Workshop, Luxembourg, Patrick.DeCausmaecker@kahosl.be
8
Technological DomainsTechnological Domains
Intelligent agents, mobile or notIntelligent agents, mobile or not Rule based systemsRule based systems Machine learningMachine learning Ontology buildingOntology building
– See related proposal “See related proposal “An ontology for planning An ontology for planning applicationsapplications” by Peter Demeester” by Peter Demeester
Data miningData mining