Developing Distributed Semantic Systems

Building Distributed Semantic Systems

Antonio Garrote Hernández

Monday, December 21, 2009

Definition

• Distributed processes

• Semantic meta data exposed as resources

• Triple space operations

• RESTful interface

Monday, December 21, 2009

Different abstraction levels

Monday, December 21, 2009

Monday, December 21, 2009

Sample application

Monday, December 21, 2009

• Sensor processes monitoring generation of exceptions in different nodes and languages/platforms

• Aggregator processes retrieving the exceptions and creating Exception resources

• Exception RESTful semantic web service

• Web client displaying generated exceptions

Exception logging system

Monday, December 21, 2009

Methodology

Monday, December 21, 2009

4 step process

1. Knowledge modeling

2. Formal description

3. Implementation of RESTful semantic services

4. Implementation of client processes

Monday, December 21, 2009

1.Knowledge modeling

Monday, December 21, 2009

1.Knowledge modeling

• Building of an ontology for the problem domain

• Ontology = TBox in Description Logics

• Use of W3C’s standard languages: RDFS or OWL

• Assistance of graphical tools like Protégé

Monday, December 21, 2009

Exception

generatedplatform

level message createdAt

ipAddress domain dc:title formatPattern

Node Language

Sample application ontology

Monday, December 21, 2009

Developing the sample application ontology with Protégé

Monday, December 21, 2009

Ontology(<http://semantic_rest/exceptions.owl>

Declaration(Class(Exception))SubClassOf(Exception owl:Thing)Declaration(Class(Node))SubClassOf(Node owl:Thing)Declaration(Class(ProgrammingLanguage))SubClassOf(ProgrammingLanguage owl:Thing)

Declaration(ObjectProperty(platform))ObjectPropertyDomain(platform Exception)ObjectPropertyRange(platform ProgrammingLanguage)

Declaration(ObjectProperty(generatedIn))ObjectPropertyDomain(generatedIn Exception)ObjectPropertyRange(generatedIn Node)

Declaration(DataProperty(level))DataPropertyDomain(level Exception)DataPropertyRange(level rdfs:Literal)

Declaration(DataProperty(ipAddress))DataPropertyDomain(ipAddress Node)DataPropertyRange(ipAddress xsd:string)

Declaration(DataProperty(formatPattern))DataPropertyDomain(formatPattern Exception)DataPropertyRange(formatPattern xsd:string)

Declaration(DataProperty(createdAt))DataPropertyDomain(createdAt Exception)DataPropertyRange(createdAt xsd:dateTime)

Declaration(DataProperty(content))DataPropertyDomain(content Exception)DataPropertyRange(content xsd:string)

Declaration(DataProperty(domain))DataPropertyDomain(domain Node)DataPropertyRange(domain rdfs:Literal))

Resulting ontologyusing OWL

Functional Notation

Monday, December 21, 2009

2.Formal description

Monday, December 21, 2009

• Formal description of the whole distributed system

• Concerned with ‘dynamic’ aspects of the system

• Use of an extended version of the Pi-Calculus

Monday, December 21, 2009

S(r)

S(j)

TEAEWGET subscribe n POST n

Formal description of the sample application

Monday, December 21, 2009

Formal description of the sample application

Processes equations:

Monday, December 21, 2009

3.Implementation of RESTful semantic

services

Monday, December 21, 2009

• 1 RESTful semantic service per OWL/RDFS class in the ontology

• Use of the Plaza Framework developed for building RESTful semantic web services

Monday, December 21, 2009

• OWL/RDFS class triples

• Exposed as a RESTful resource

• Conceptually equivalent to a Linda triple space

RESTful Semantic Web Service

Monday, December 21, 2009

Service’s InterfaceCreate POST

Read GET

Blocking read GET + xblocking header

Subscribe • GET + xsubscribe header• Websockets

Update PUT

Destroy DELETE

Monday, December 21, 2009

resourcetriplespace

REST Resource

HTTP request

triple space op.

RDF Graph

SPARQL query

Lifting

Lowering

Monday, December 21, 2009

Plaza Framework

Monday, December 21, 2009

Plaza Framework

• Manages ‘Plaza Applications’: OWL ontolgy + configuration module + resources modules

• Declarative: almost no coding required

• Support for different Java RDF repositories

• Automatic managing of triples through the concepts of write-tree and read-tree for the resources

• Support for the subscribe operation through chunked HTTP responses and websockets

Monday, December 21, 2009

Sample application configuration module

Monday, December 21, 2009

Sample application exceptions resource module

Monday, December 21, 2009

4.Implementation of client processes

Monday, December 21, 2009

• Client processes are just REST clients

• Required libraries: HTTP and triples format parser (RDF/XML, Turtle, JSON)

• Optional support for Websockets

• Clojure library included as part of the Plaza Framework

Monday, December 21, 2009

Clojure implementation of the sensor and aggregator processes

Monday, December 21, 2009

Javascript web client using websockets

Monday, December 21, 2009

Web client execution

Monday, December 21, 2009

• Semantic technologies enable rich description of problem domains vs. relational modeling

• Formal description of distributed systems enables reasoning about its validity and correctness

• RESTful semantic web services enable easy access to semantic data

• Building RESTful semantic web services clients is trivial due to the use of well known HTTP standard

• Triple Space operations are powerful coordination mechanisms for distributed processes

Conclusions

Monday, December 21, 2009