Ambient Intelligence through Ontologies Vassileios Tsetsos [email protected] P-comp Research Group .

Ambient Intelligence through Ontologies

Vassileios Tsetsos [email protected]

P-comp Research Grouphttp://p-comp.di.uoa.gr

What is an ontology?

A formal, explicit specification of a shared conceptualization. (Studer 1998, original definition by Gruber in 1993)

Formal: it is machine-readable Explicit specification: it explicitly defines concepts,

relations, attributes and constraints Shared: it is accepted by a group Conceptualization: an abstract model of a phenomenon

What is an ontology?

Taxonomy, classification, vocabulary, logical theory, … Concepts/classes, relations, properties/slots,

instances/objects, restrictions/constraints, axioms, rules

Heavyweight vs. Lightweight

They differ in expressiveness, reasoning capabilities, complexity, decidability.

Lightweight E-R diagrams, UML

Heavyweight Description Logics, frames, first order logic

There are W3C standards for each case (RDF, RDF Schema, OWL)

We should choose carefully!

Types of Ontologies (1)

Upper Level OntologiesDescribe very general concepts. SUO (IEEE Standard Upper Ontology)

KR OntologiesRepresentation primitives => Semantically-

described grammars of ontology languages. OKBC, OWL KR, RDF Schema KR

Types of Ontologies (2)

Domain Ontologies Are specializations of Upper Level Ontologies,

reusable in a given domain (e.g., a generic ontology for smart environments)

Unified Medical Language System (UMLS)

Application Ontologies They model all the knowledge required for a particular

application (e.g., an ontology for a specific smart classroom)

Some examples

IEEE SUO

RDF(S) KR

Many advantages

Provide formal model descriptions that allow reasoning They support common queries:

Queries about the truth of statements (Is there a printer in room I9?) Queries expecting an object to be returned (Where is John?)

Are quite scalable (especially Semantic Web ones) Provide interoperability as they are agreed by a community

(…at least this should be the case!) SW ontology languages

are XML-based => XML advantages have been standardized and are widely used

…

Pervasive Computing (PC)

Computing paradigm that envisages: Ubiquitous networking and service access Intelligence Intuitive HCI Context-awareness Seamless interoperation between heterogeneous

agents Privacy and Security …

Ontology applications in PC

Context modeling & reasoning Context ontologies (location, time) which define

structure and properties of contextual information Semantic Web Services

Semantic description => automated discovery and matchmaking, composition, invocation, …

Semantic interoperability between heterogeneous systems (e.g., agents) through a shared set of concepts

Security and trust

Some “PC+Ontologies” projects

CoBrA SOUPA Gaia Other

CoBrA (1)

eBiquity Research Group, UMBC http://ebiquity.umbc.edu

A broker-centric agent architecture that aims to reduce the cost and difficulties in building pervasive context-aware systems.

In this architecture, a Context Broker is responsible to: Acquire & maintain contexts on the behalf of resource-poor

devices & agents Enable agents to contribute to and access a shared model of

contexts Allow users to use policy to control the access of their personal

information

CoBrA (2) Context Broker:

maintains a model of the present context and shares this model of context knowledge with other agents, services and devices.

CoBrA ontologies

A set of ontologies that specialize the SOUPA Ontology.

They model the context and the processes of pervasive environments.

E.g., CoBrA Place models different types of “Place” on a

university campus

CoBrA Place Ontology

SOUPA (1)

Standard Ontology for Ubiquitous and Pervasive Applications (SOUPA)eBiquity @ UMBC,

http://pervasive.semanticweb.orgWritten in OWL

SOUPA (2)

Gaia (1)

A PC infrastructure for smart spaces CORBA-based middleware for the

management of Spaces Ontologies written in DAML+OIL

Gaia (2)

Ontology Server: definitions of terms, descriptions of agents and meta-information about context available in a Space

Checks ontology consistency and provides maintenance

Semantic interoperability is performed through the common adoption of the same ontologies by all agents

Ontologies also help the developer to write inference rules or machine learning code in a generic way

Other uses of ontologies in Gaia

Configuration management New unknown entities may enter a Space In earlier version: scripts & ad hoc configuration files

Semantic discovery with a FaCT Server Semantic queries involve subsumption and classification of

concepts Context modeling

Context is modeled as predicates e.g., temperature (room3,”-”,98F) Ontologies describe the type and values of predicate arguments

Context-sensitive behavior The developers can specify the behavior of the applications

under certain contextual conditions through the supported ontologies.

The Gaia infrastructure

Gaia context infrastructure

The ontology infrastructure of Gaia

CONON: The context ontology

Extensible ontology comprised of: Upper Level Ontology Specific Ontology

Written in OWL Enables DL reasoning (subsumption,

consistency, instance checking, implicit context from explicit context) with OWL-Lite axioms

Enables First Order Logic reasoning (inference of higher level context) with user-defined rules

Trust

SW entails a Web of Trust PC requires ad-hoc soft-security models Ontologies can model semantic networks

of trusted entities and allow trust inference Ontologies are used for the definition of

(rule-based) Policy Languages Rei, KAoS

Trust inference

Directly connected nodes have known trust values

Trust for not directly connected nodes can be inferred with several algorithms: Maximum and minimum capacity paths (~ the range

of trust given by neighbors of X to Y) Maximum and minimum length paths (~ how “far” is Y

from X?) Weighted average (~ recommended trust value for X

to Y). It is a very complex algorithm!!! Why?

Complexity of trust computation

Trust is affected by social, contextual and other ad hoc conditions

Example (on the subject of “AutoRepair”) A distrusts B, B distrusts C => A trusts C?

A may want to trust C, because B distrusts C If C cannot be trusted by B, A may distrust C even more

A complete solution: semantic descriptions of trusted entities and user-defined trust policies

FOAF Ontology

Builds social networks Individuals are described by name, e-mail, homepage, etc. There are links between individuals

A trust ontology (1)

Nine levels of trust (trustsHighly, distrustsSlightly, etc.)

Extending foaf:Person (1)<Person rdf:ID="Joe">

<mbox rdf:resource="mailto:[email protected]"/>

<trustsHighly rdf:resource="#Sue"/>

</Person>

A trust ontology (2)

Extending foaf:Person (2)<Person rdf:ID="Bob">

<mbox rdf:resource="mailto:[email protected]"/><trustsHighlyRe>

<TrustsRegarding> <trustsPerson rdf:resource="#Dan"/> <trustsOnSubject

rdf:resource="http://example.com/ont#Research"/></TrustsRegarding>

</trustsHighlyRe><distrustsAbsolutelyRe>

<TrustsRegarding> <trustsPerson rdf:resource="#Dan"/> <trustsOnSubject

rdf:resource="http://example.com/ont#AutoRepair"/></TrustsRegarding>

</distrustsAbsolutelyRe></Person>

Current and future work in P-comp

Semantic Web Services Description Logics Location modeling Tools survey and experimentation Meta-information for sensor data Ontologies for medical applications Any ideas???

Location modeling (1)

Ontologies can map and interconnect different underlying spatial representations

This facilitates advanced reasoning and user-defined queries

A “location modeling team” is currently being formed to design and develop a system: With human-centered, 3D indoor spatial representation Which supports declarative and semantically-rich queries Which supports mobile users and location prediction Which seamlessly integrates different spatial representation

approaches (set-based, graph-based, geometric)

Location modeling (2)Top-Level

Location Ontology

ApplicationOntology 1

ApplicationOntology 2

ApplicationOntology 3

Oracle Spatial

DOMINO LocationOntology

Repository

Model Mapping Engine 1

Model MappingEngine 2

Model Mapping Engine 3

ExplicitSemantics

User Applications(e.g., navigation)

QueriesThis is actually a Domain Ontology

(Prediction-driven) Events

Different DB platforms,

access terms, conceptual

models

Some open research issues

Can they efficiently model sensor data? Will the introduction of Probability elements

improve their effectiveness? If yes, how can this be implemented?

Development of user-friendly tools and powerful & efficient reasoners

Automated ontology generation/extraction and easy ontology maintenance

Further reading Ontological Engineering, Gómez-Pérez, Fernández-López, Corcho, 2004,

Springer Harry Chen et al., "SOUPA: Standard Ontology for Ubiquitous and

Pervasive Applications", International Conference on Mobile and Ubiquitous Systems: Networking and Services, August 2004.

Harry Chen et al., "A Context Broker for Building Smart Meeting Rooms", Proceedings of the Knowledge Representation and Ontology for Autonomous Systems Symposium, 2004 AAAI Spring Symposium, March 2004.

Robert E. McGrath, Anand Ranganathan, Roy H. Campbell and M. Dennis Mickunas, Use of Ontologies in Pervasive Computing Environments

Xiao Hang Wang, et al., Ontology Based Context Modeling and Reasoning using OWL, Second IEEE Annual Conference on Pervasive Computing and Communications Workshops, 2004

Jennifer Golbeck, James Hendler, Trust Networks on the Semantic Web, WWW 2003

RDFWeb: FOAF: ‘the friend of a friend vocabulary’, http://rdfweb.org/foaf/