network ontology

33
Network Ontology Ramesh Subbaraman Soumya Sen UPENN, TCOM 799

Upload: zan

Post on 06-Jan-2016

30 views

Category:

Documents


1 download

DESCRIPTION

Network Ontology. Ramesh Subbaraman Soumya Sen UPENN, TCOM 799. Motivation. Semantic web – i.e. make web machine UNDERSTANDABLE. Ontology: vocabulary - semantics - relationships - rules. Resource Description Framework – RDF. - resources - properties - statements. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Network Ontology

Network Ontology

Ramesh SubbaramanSoumya SenUPENN, TCOM 799

Page 2: Network Ontology

Motivation

Semantic web – i.e. make web machine UNDERSTANDABLE.

Ontology: vocabulary- semantics- relationships- rules

Page 3: Network Ontology

Resource Description Framework – RDF

- resources- properties- statements

Page 4: Network Ontology

Resource Description Framework Schema – RDFS

A schema language to model the resources

Page 5: Network Ontology

Partial Example

Page 6: Network Ontology

Axioms in RDF(S)

We have that:

so, in RDF we add a resource

where symmetric is agreed to mean

composition

Page 7: Network Ontology

General Axiom

Page 8: Network Ontology

Why Network Ontology?

Ontology is the shared understanding of what concepts or specifications mean in a particular domain.

Presently there isn’t any formal widespread ontology that is being used to represent networks (some proprietary ones exist).

A formal ontology will make creating, updating, and simulating networks more efficient.

Page 9: Network Ontology

Possible solution:

Develop a Network Markup language used to represent all information needed in a network.

By creating an XML-based ontology it will provide greater flexibility when dealing with platforms and programming.

NML has been designed with a top-down approach. Identification of ‘classes’ and establish relationship

between them for different scenarios. Validating relationship with Ontology languages.

Page 10: Network Ontology

Today’s topics:

Network markup language for Network Planning tool.

Framework for modeling wireless networks. Ontology for modeling Sensor Networks. Ontology of representation of sensor network

data. Semantic services for sensor-rich Information

systems.

Page 11: Network Ontology

Network Planning Tools

Expert systems like CACIT, Opnet are good but they have some drawbacks. So new expert systems that may be developed can use ontology based on NML for representing the network layout.

Kinds of questions asked to the user: -What kind of company? -How many branches? Regional/HQ/global-Workload estimate-description of office dimension

Page 12: Network Ontology

Output of NPT:

Output:

-Well illustrated network layout diagram

-satellite/LAN/local loops/wireless/bluetooth

-network plans for each office and each floor, access points, routers.

-future extension capabilities etc.

Page 13: Network Ontology

Idea of using XML ontology:

Describing and identifying the commonly used patterns in a network and capturing them in the XML ontology.

The representation through relations and tags help in easy interpretation by automated processes.

Page 14: Network Ontology

NML DTD

A DTD was created to be used in the digital network advisor tool and captures the necessary parameters to fully describe all types of networks. The fields are determined by the user’s answers, underlying assumptions, and calculations.

Page 15: Network Ontology

NML DTD

Page 16: Network Ontology

XML output

Page 17: Network Ontology

Graphical representation of data extracted from XML ontology

Page 18: Network Ontology

Proposed ontology for Wireless networks framework:

Page 19: Network Ontology

“A framework for modeling sensor networks”- Raja Jurdak, C. Lopes, P.BAldi, OOPSLA,’04.

Proposes a framework for sensor network modeling based on general features identified through analysis of existing sensor networks.

The framework facilitates the modeling of new sensor networks by characterizing them according to these general features and providing a set of performance goals.

Specification of network performance requirements helps in selecting communication protocols.

Related Literature: “Ontology driven Adaptive-sensor networks”, S.Avancha, C.Patel, A. Joshi, UMBC.

Page 20: Network Ontology

Elements in the framework of sensor networks:

Topology- affects network routing, power consumption, battery life. Includes physical (deployment area) and logical organization of the network as well as sensor density.

Sensor networks can have distributed or a clustered organization where selected nodes handle forwarding.

Page 21: Network Ontology

Elements in the framework of sensor networks (cont..):

Network Setting- communication media quality (noise, spreading, attenuation, multi-path fading) and operating environment (security).

Sensor Description- resources, memory, battery, participation, processing speed.

Data flow- event-driven/demand-driven (SQL type query), processing architecture, system health.

Page 22: Network Ontology

Proposed ontology for Sensor Networks:

Page 23: Network Ontology

“A Novel Ontology for Sensor Networks Data”- M.Eid, R.Liscano, A. El Saddik, MCRLab

Sensor data from a large number of nodes need to be searched efficiently. Classical retrieval techniques had poor performance.

Brings improvement in search engine performance by utilizing captured relationships.

This ontology is based on the IEEE 1451.4 smart transducers template description language.

Page 24: Network Ontology

Components of Ontology:

Classes or concepts that may have sub-classes.

Properties or relationships that describe various features and properties of the concepts (also called slots).

Restrictions on slots that are superimposed on the defined classes and/or properties to define allowed values (domain & range).

Page 25: Network Ontology

1. Identification of the initial taxonomy:Step: List of concepts as described by the identified terms and form the initial class taxonomy.

Page 26: Network Ontology

2. Properties & restrictions:

Relationships among classes are usually referred to as properties.

Property links an individual from its domain to an individual of its range.

Concepts can be refined by superimposing constraints and axioms expression relationships.

Universal restriction & Existential restriction.

Page 27: Network Ontology

2. Properties & restrictions: (contd..)

Page 28: Network Ontology

3. Consistency Checking:

2 types of Ontology Tests:

- Subsumption test (Testing Class hierarchy)

- Consistency check (logical check)

Usage of possible Ontology development tools: Protégé. Knowledge representation language for modeling various data types of sensor data is OWL-DL. Validation tool: RacerPro.

Page 29: Network Ontology

“Towards Semantic Services for Sensor-rich Information Systems”- J.Liu, F.Zhao, Microsoft Research (publ. IEEE 2005)

Need to describe the architecture of a semantic-service-oriented sensor information system platform.

The key to enabling scalable sensor is to define an ontology and associated information hierarchy for interpretation of raw data streams.

Systems may have mobile or stationary nodes, and products from different vendors must interoperate.

Related literature: “Semantic Agent Technologies for tactical sensor networks”, G.Jiang, W.Chung, G.Cybenko

Page 30: Network Ontology

Application in a garage parking sensor system

Page 31: Network Ontology

Application in a garage parking sensor system composed of semantic services

Page 32: Network Ontology

Advantages of Ontology designs

Sharing common understanding of the structure of information among people or software agents.

Enabling reuse of domain knowledge. Making domain assumptions explicit. Separating the domain knowledge from the

operational knowledge Analyzing domain knowledge.

Page 33: Network Ontology