Semantic Web-based Knowledge Management in Distributed Systems

CANS 2008

Dr. Sabin‐Corneliu Buraga www.purl.org/net/busaco

Semantic Web‐based Knowledge Management in Distributed Systems

Sabin C. Buraga“A. I. Cuza” University of Iaşi, Romania

Faculty of Computer Science

CANS 2008


Outline

GoalSemantic Web – challenges & realities

Knowledge managementCase studiesConclusions

CANS 2008


Goal

Using semantic Web to model knowledgewithin distributed systems

distributed file systems (Buraga, LNCS, Springer, 2002)multi‐agent systems (Hritcu & Buraga, IEEE, 2005;

Buraga, Rusu & Cioca, CRC Press, 2008)Web services (Buraga & Gabureanu, IEEE, 2003;

Buraga & Rusu, Elsevier, 2006; Brut & Buraga, IEEE, 2008)Grid computing (Alboaie & Buraga, IEEE, 2003;

Brut & Buraga, IEEE, 2008)collaborative Web applications (Dumitriu & Buraga, IEEE, 2007; Dumitriu, Girdea & Buraga, AAAI, 2007; Luca & Buraga, ICEIS, 2008)

CANS 2008


Realities

Information overload

Same old keyword‐based search

Identity abuse the need for social verification

Web applications are still rigid:each site has got its data and it is not sharing it

CANS 2008


Realities

computers can not understand anything

CANS 2008


Necessity

attaching metadata to Web resources

vocabularies describing “things”:properties, domains, persons,…

CANS 2008


Necessity

specifying relations between resources

CANS 2008


Necessity

managing knowledge about things

CANS 2008


Necessity

The implicit knowledge must be explicitly specified

“java” ≡ language, island, or coffee?

CANS 2008


Necessity

The implicit knowledge must be explicitly specified

Java is a programming languageCANS is a workshop organized by the University of Tg. Mureş

Distributed systems can be complex

Statements that can be figured out by (some) people…But the computers can understand them?

CANS 2008


Solution

RDF (Resource Description Framework)

attach metadata and specify relations between resources

a more flexible model than relational model of databases

important brick of the semantic Web

CANS 2008


RDF

RDF model is based on triples

entity has property with value

URI/IRIURI/IRI

URI/IRIliteral

CANS 2008


RDF – examples

:Java rdf:type :Language:CANS rdf:type :Workshop

:CANS :organisedBy :TgMuresUniv:DistribSystem rdfs:subClassOf :ComplexSystem

:busaco :hasName "Sabin Buraga"

using (standard) vocabularies

CANS 2008


Semantic mash‐ups via data repositories: Linked Open Data

CANS 2008


Ontologies

The need of modeling classes of resources andcomplex relations/restrictions between them

CANS 2008


Knowledge about resources can be sharedwithin a given community of practice

structuring information conform to different points of view

AAA – Anyone can say Anything about Any topic

Ontologies

CANS 2008


Ontologies

Taxonomies

ACM, North American Industry Classification System, Open Directory, Yahoo!

CANS 2008


Ontologies

Thesauri

AGROVOC, Center for Army Lessons Learned Thesaurus,National Cancer Institute Thesaurus, WordNet

CANS 2008


Ontologies

Complex ontologies

Gene Ontology, OWL‐S: Process Ontology,Transportation Ontology, Web Service Modeling Ontology,

Wine Ontology

IEEE Standard Upper Ontology, OpenCyC

CANS 2008


Ontologies

Financial ontology – excerpt (Amit Sheth, 2006)

CANS 2008


Ontologies

Ontology editing tools ontology (Buraga, Cojocaru & Nichifor, 2006)

CANS 2008


expressed by standardized languagesOWL (Web Ontology Language)

Ontologies

CANS 2008


Class (participant intersectionOf (student young person))Class (participantintersectionOf (restriction (hasPet allValuesFrom (penguin))restriction (hasPet someValuesFrom (animal))))

Ontologies – example

Every participant must have at least one penguin,because her/his has a pet and all pets must be penguins

Inference

CANS 2008


ObjectProperty (hasPet domain (person) range (animal))Individual (Alice type (young) type (student)

value (hasPet Tux))

Ontologies – example

Alicemust be a person – owners of pets are persons –and she is a participant

Tuxmust be a penguin (all pets of participants are penguins)

CANS 2008


Using these statements, Web applications can reason

the need of specifying rules

if P is a participant, then P is paying attentionsome participants are bright

Beyond ontologies

support for making decisions

CANS 2008


Semantic Web

Layers of the Semantic Web (Tim Berners‐Lee, 2006)

CANS 2008


Semantic Web vs. Distributed Systems

How semantic Web technologies could be usedto model knowledge within distributed systems?

distributed file systemsmulti‐agent systemsWeb/Grid services

collaborative Web applications: wikis…

CANS 2008


Knowledge Management Issues

Advanced Knowledge Technologies (AKT) manifesto:6 challenges concerning the engineering and

management of knowledge: acquiring modelingreusingretrievingpublishingmaintaining

CANS 2008



The system should acquire knowledge, and not formatted text – provided by the end‐users

A suitable model to store knowledge is a triple‐based one(RDF model)

Semantic markups – embedding semantics within Web pages microformats, RDFa etc.

CANS 2008



Information gathering involves obtaining selected (semi)structured data from external sources

RSS/Atom feeds, microformats, Web services,CGI scripts, RDF stores, databases

examples: GRDDL (Gleaning Resource Descriptionsfrom Dialects of Languages), Triplify, Yahoo! Search Monkey

CANS 2008



Knowledge modeling

although each piece of information should be semantically modeled, this approach seems almost impossible in practice

CANS 2008



Knowledge modeling

certain important information can be expressed by following the object‐oriented paradigm

CANS 2008



Knowledge modeling

classes needed to be used are identified

properties are defined

information regarding the individuals (class instances)is filled in via collaborative mechanisms by the involved users

CANS 2008



Knowledge modeling

information can be properly organizedfor both human and computer access

information of interest can be rendered according tothe user needs and preferences

CANS 2008


Case Studies

CANS 2008


Grid Computing

Sharing, selection, and aggregation of world‐wide distributed heterogeneous resources for solving

large‐scale problems in different areas of interest or for proving access to massive repositories of data,

information, or knowledge

(Buyya, 2002; Abbas, 2004; Wells, 2008)

CANS 2008


Grid Computing

Grid resources – hardware, software, logical –do not involve a particular central location

for their management

Simultaneous use of large number of resources

CANS 2008


Grid Computing

Related technologies:peer‐to‐peer architectures

cluster computingInternet & Web computing

cloud computing

CANS 2008


Grid Computing

One difficulty is a coherent management of the Grid resources

we need to adopt a knowledge‐based methodology

very useful in the context of Semantic Web

CANS 2008


Semantic Grid

We propose to use this approach for modelingthe Grid services

using semantic Web‐based descriptions for Grid services,the applications will automatically discover, invoke and

compose the desired services

inter‐operability & execution monitoring are also possible

CANS 2008


Semantic Grid

Example #1: attaching metadata to a Grid resource accessed via a Web portal

XFiles (Buraga, LNCS, 2002) – a RDF/XML vocabularyfor specifying metadata regarding distributed file systems

we also adopt well‐known metadata standards:EXIF, DCMI (Dublin Core Metadata Initiative),…

CANS 2008


Semantic Grid

Example #2: creating social networks of existing users

FOAF (Friend Of A Friend) vocabulary

<foaf:Person rdf:ID="jsbach"><foaf:name>Johann Sebastian Bach</foaf:name><foaf:givenname>Johann</foaf:givenname><foaf:family_name>Bach</foaf:family_name><foaf:homepage rdf:resource="http://www.music.org/bach"/><foaf:knows rdf:resource="#vivaldi" />

</foaf:Person>

Query via SPARQL

CANS 2008


Semantic Grid Services

Grid services can be semantically enriched by metadata and ontological descriptions

Semantic Web Services Framework (SWSF)OWL‐S and Semantic Web Services Ontology

process ontologies for Web services

Web Service Modeling Ontology (WSMO)Fensel et al., 2007

CANS 2008



Example #3: specifying a Grid service that offer access to the metadata attached to a given resource

via WSMO declarations

we define basic operations regarding the resource storage, considered as files

we can easily classify the resources by grouping them on directories – a straightforward taxonomy

CANS 2008




concepts (classes): file, owner, and directory

:File rdf:type owl:Class:Owner rdf:type owl:Class

:Owner owl:disjointWith :File…

CANS 2008




relation: ownership

:ownership rdf:type rdf:Property:ownership rdfs:domain :File:ownership rdfs:range :Owner

CANS 2008




axiom: restricts an owner to be effectively memberof the owner class

axiom ownershipFromOwnerdefinedBy ownership (?x, ?y) :− ?x [ ownerOf hasValue ?y ] memberOf owner .

CANS 2008



A possible instance could be:

instance bootstrapGlobusFile memberOf Filename hasValue ”bootstrap.jar”hasOwner hasValue root

CANS 2008


Enterprise Collaborative Systems

A mechanism of knowledge managementmust be provided by every collaborative Web system

a modern wiki must support user collaborative tools and must allow attaching metadata to the concepts and relations

established between the involved concepts

CANS 2008



As a testbed, we choose XWiki

a second generation enterprise wikibuilt on open technologies (Java)

architectural details in (Dumitriu, Girdea & Buraga, 2007)

www.wxiki.org

CANS 2008



The knowledge can be acquired via Web forms

the system gives the possibility to define classes, properties, restrictions, etc. conforming to the RDF model

CANS 2008



CANS 2008



Extending XWiki to give support for microformats(Dumitriu, Girdea & Buraga, AAAI, 2007)

generating standard microformats for each wiki document: persons (hCard), events (hEvent),

reviews (hReview), tags (rel‐tag) etc.

defining a new microformat (hLocation) to specify geographical locations in fuzzy terms: near, in front of,…

CANS 2008



We can model relations regarding the proximity of certain points of interest for a given user/group

CANS 2008



A non‐intrusive microformats‐based recommender agent (Luca & Buraga, ICEIS, 2008) is also provided

understands the behavior of a user on the Web and filters out the irrelevant data, presenting to the useronly the information he/she is most interested in

machine learning, Java, JavaScript, native XML database

developed as a Firefox extension

CANS 2008


CANS 2008


Conclusions

There are certain important challenges that concern the engineering and management of knowledge

within complex distributed systems

One possible solution is to adopt the actual Semantic Web technologies – metadata and ontological constructs –for specifying the involved knowledge and processes

CANS 2008


Conclusions

It’s not the wires – it’s the computers

It’s not the computers – it’s the documents

It’s not the documents – it’s the things

“The last level of abstraction is the Web of real things,built on top of the Web of documents,

which is in turn built on the network of computers.”Tim Berners‐Lee, 2008

CANS 2008


Summary

Semantic Web‐based modeling of resources– using microformats, metadata, and ontologies –

within distributed systems:Grid applications and enterprise wikis

CANS 2008


Questions?

Thank you!

Technology

Semantic Web-based Knowledge Management in Distributed Systems