Semantic modeling of System Requirements

1
Semantic modeling of System Requirements

• Lunch presentation _at_GDMC
• Lieke Verhelst MSc Student for GIMA
• Feb 27th 2009

2
Problem Area
Figure taken from "OGC Sensor Web Enablement
Overview And High Level Architecture", Mike
Botts, George Percivall, Carl Reed, John Davidson
Eds. OGC, 2007
3
Examples of ST sensor data

• Recordings taken from radiosonde and pilot
  balloon observations, stored in the Integrated
  Global Radiosonde Archive (IGRA).
• Records from the Antarctic Iceberg Tracking
  Database
• GPS track (Garmin)

4
User questions

• How did the frost line move over the years?
• Where and when do icebergs a and b collide?
• Do bikers and hikers get in each other's way in
  area x?
• Combinations of
• Time
• instants or intervals
• historic data or real data
• Space
• Point, line, region

5
Database implementation

• Depends on
• Data set (real time, historic data)
• User question (spatial analysis? temporal
  analysis?)
• System requirements (speed, storage)
• Available solutions (Oracle, DB2, Informix,
  Postgres/GIS, ESRI)
• Involves
• Information analysis (data type, data model)
• DBA work (indexes, storage)

6
MSc Research

• Objective to provide users of sensor data with
  guidelines how to choose the right database
  implementation for their purpose. If possible
  automate this process.

7
Research question

• Designing a method that models the concepts of
  sensor data (e.g. sample frequency, size of data
  file, structure of sensor data) in order to
  generate a suitable database implementation
• Apply a weighted qualification to the proposed
  result
• Automate the process to the DB implementation

8
How can we model this?

• UML
• Via Classes to tables
• Via stereotypes and tagged values to index types
• From UML to DDL
• but how to (automatically) guide the user
  towards selection of the appropriate
  implementation? -gt expert system

9
Semantic modeling
Index
SolutionFor
Speed
lt?xml version'1.0' encoding'UTF-8'?gt lt!DOCTYPE
rdfRDF lt!ENTITY rdf 'http//www.w3.org/1999/0
2/22-rdf-syntax-ns'gt lt!ENTITY kb
'http//protege.stanford.edu/kb'gt lt!ENTITY
rdfs 'http//www.w3.org/2000/01/rdf-schema'gt gt lt
rdfRDF xmlnsrdf"rdf" xmlnskb"kb"
xmlnsrdfs"rdfs"gt ltrdfsClass
rdfabout"kbIndex" rdfslabel"Index"gt ltrdfs
subClassOf rdfresource"rdfsResource"/gt lt/rdfs
Classgt ltrdfsClass rdfabout"kbSpeed"
rdfslabel"Speed"gt ltrdfssubClassOf
rdfresource"rdfsResource"/gt lt/rdfsClassgt ltrdf
Property rdfabout"kbsolutionFor"
rdfslabel"solutionFor"gt ltrdfsdomain
rdfresource"kbIndex"/gt ltrdfsrange
rdfresource"rdfsClass"/gt lt/rdfPropertygt lt/rdf
RDFgt
10
OWL

• extended RDF
• Constraints, relationships among resources,
  cardinality, domain and range restrictions,
  union, disjunction, inverse, transitive
• OWLDL based on description logic
• existential quantifier ?, which can be read as
  at least one, or some
• universal quantifier ?, which can be read as
  only
• rules (!!)

11
Research work Steps

• Create OWL-DL ontology
• Apply logic rules
• Generate DDL
• Implement in DB

12
Software Used

• Create OWL-DL ontology (Protégé)
• Apply logic rules (Pellet or Racer, Jess,
  JessTab)
• Generate DDL (OWL-gtUML-gtDDL, Poseidon, Eclipse)
• Implement in DB

13
DEMOS
1. Demonstrate how reasoner can categorize
instance to predefined class
2. Demonstrate use of rule engine