Building a Database Semanticizer for Task Computing

1
Building a Database Semanticizer for Task
Computing

• Anubhav Sonthalia
• MS Thesis Defense, 12/06/04

2
Overview

• Basics
• Introduction
• Task Computing
• Database Semanticizer
• BioGrid Task Computing An application
• Conclusion
• Demo

3
What is the Semantic Web

• The Semantic Web is an extension of the current
  Web, in which data is given a well-defined
  meaning by representing it in RDF and linking it
  to commonly accepted ontologies. Tim Berners
  Lee, October 2002

4
Task Computing

• Our answer for ubiquitous computing challenges
  based on the Semantic Web

5
What are Semantics A Task Computing Definition

• Semantics is a set of clues and glues for
  resources (information, services) to be used in
  different systems (computers and humans) and
  contexts.
• Semantics is
• Interoperability-enabler/enhancer
• Measured by extent of interoperability it enables

6
Problem

• Most Structured data is stored in relational
  databases.
• We need to map it to OWL in order to use it in
  the semantic context.

7
Solution

• Developed a semi automatic approach to create a
  map between relational databases and semantic
  objects.

8
Goal of Task Computing

• User wants to do Tasks
• Services offer means
• Web services, UPnP services etc.

Filling the gap
9
Task Computing - Definition

• is a user-oriented framework that lets non-expert
  users accomplish complex tasks in application-,
  device-, and service-rich environments.
• provides a myriad of ways for users to interact
  with and through ubiquitous environments.

10
TC in eHome
11
TC in a Car

• GPS Navigator (or PDA) to control

• TC Voice Menu
• Do you want to
• Navigate to an address in PDA
• Play Music from Digital Audio Player

12
TC in an Office
13
TC at an Airport
14
TC at a Coffee Shop
15
Task Computing Technology

• Major components are
• Service Discovery
• Service Composition
• Service Execution
• Service Publishing

16
Task Computing Architecture
17
Grasping the Environment

• Discover services through UPnP
• Acquire Semantic Service Description (SSD) using
  UPnP requests.
• Business card metaphor
• With a business card (SSD), you can put the
  person (service) in a social context (Web
  ontology)
• Then you can interact with the person (service)

18
Semantic Service Compositions
Addr Address Cont Contact WP Web Page
Discovered Services
19
Task Execution
My Contact OWL-S
Business Address of OWL-S
Route from FLA to OWL-S
Contact
Address
Grounding
Grounding
Grounding
STEER
Web Service Call with no input
Result (OWL String)
UPnP Call with a string
Result (None)
My Contact Web Service
Route from FLA to UPnP
Route from FLA to Business Address of My Contact
20
TC Service Execution (Instance View)
Semantic Contact Instance
Semantic Address Instance
ltrdfRDF gt ltAddress gt ltCountrygtUSAlt/Countrygt
ltStreetAddressgt8400 Baltimore Ave
lt/StreetAddressgt ltCitygtCollege Parklt/Citygt
ltStategtMDlt/Stategt ltZipCodegt20740lt/ZipCodegt
lt/Addressgt lt/rdfRDFgt
ltrdfRDF gt ltContact rdfID"Contact_Ryusuke_
Masuoka"gt lthasBusinessAddressgt ltAddressgt
ltCountrygtUSAlt/Countrygt ltStreetAddressgt8400
Baltimore Ave lt/StreetAddressgt
ltCitygtCollege Parklt/Citygt ltStategtMDlt/Stategt
ltZipCodegt20740ltZipCodegt lt/Addressgt
lthasBusinessAddressgt lthasHomeAddressgt
lt/hasHomeAddressgt ltFirstNamegtRyusukelt/coFirstNam
egt ltLastNamegtMasuokaltLastNamegt lt/Contact gt
lt/rdfRDFgt
ltsoapenvEnvelope gt ltsoapenvBodygt ltGetWeather
gt ltzip type"xsdstring" gt 20740lt/zipgt
lt/GetWeathergt lt/soapenvBodygt lt/soapenvEnvelope
gt
Weather Info of Web Service
ltsoapenvEnvelope gt ltsoapenvBodygt ltGetPhoto
gt ltstreet type"xsdstring" gt 8400
Baltimore Avelt/streetgt ltcity
type"xsdstring" .gt College Parklt/citygt
ltstate type"xsdstring" gtMDlt/stategt ltzip
type"xsdstring" gt20740lt/zipgt lt/GetPhotogt
lt/soapenvBodygt lt/soapenvEnvelopegt
Aerial Photo of Web Service
ltsoapenvEnvelope gt ltsoapenvBodygt
ltInitService gt ltaddress type"xsdstring"
gt 8400 Baltimore Ave, College Park, MD
20740 lt/addressgt lt/InitServicegt
lt/soapenvBodygt lt/soapenvEnvelopegt
Route from FLACP UPnP Service
21
Service Publishing

• Lets the user publish services from OS/App/OWL
  objects, and OWL-S
• Publish service Make services OWL-S through
  UPnP
• Numerous ways of publishing services
• White hole,
• bank service,
• multimedia publisher,
• image generator, etc.

22
White Hole and PIPE
OS/Application objects
Semantic objects
Semantic Service Description
Semantic -ization
NoProcessing
NoProcessing
White Hole
Semantic objects
Web Service API
Create a web service with WSDL to publish the
object instance, and generate an OWL-S file to
describe the web service
NoProcessing
PIPE
OWL-S service description with grounding to web
service
Publish the OWL-S description file
PIPE - A Service Management Tool
Allow Task Computing Client to discover
23
Revisit the Problem

• Most Structured data is stored in relational
  databases.
• We need to map it to OWL in order to use it in
  the semantic context.
• Important in Bio-Grid Task Computing domain

24
Database Semanticizer

• We have created a tool which enables easy
  mappings between the databases and commonly
  accepted ontologies.
• It creates the semantic instance objects.
• It can publish the object as a service in order
  to enable us to pick it up from TCC.

25
Process Overview
26
Modules

• Database Web Service Wrapper
• Application
• Service and Control UI.

27
System Architecture
28
Web Service Wrapper

• Connects to a remote database or a remote Grid.
• Uses the OLE DB interface.
• In grid environments, we may not be able to
  directly connect to a database. Hence this
  wrapper is a way of getting around the problem.

29
contd

• Provides various Web methods
• GetTables
• GetTableDetails
• ExecuteNonQuery
• ExecuteQuery

30
Application Body
31
Contd

• GUI Interface
• Displays the various Tables and Columns with
  details. Helps the user grasp the schema of the
  database before he makes any SQL queries
• Execute SQL queries to reach the final object
  tuples we want to use and convert to semantic
  objects
• Enter Ontology URI using which semantic objects
  need to be created.

32
Mapping Editor
33
Mapping Editor

• Parses the ontology and creates a tree view of
  the objects in the ontology.
• Also shows the various datatype and object
  properties.
• Shows the various attribute elements for the
  selected SQL query.
• User can create a mapping using simple clicks.
• Save these mappings if desired.
• Use saved mapping.

34
Contd

• The mapping editor has been written using C.
• There is no OWL parser for C.
• C OWL Parser built on top of the Drive RDF
  Parser from CMU.
• OWL Parser provides us a programmatic API to
  parse OWL ontologies.

35
Service Control UI

• Web page interface
• giving the user control over the application
• or accepting user input.
• In our case,
• displays the query results in the form of a table
  
• allows the user pickup an object.
• write semantic object file,
• publish through PIPE
• pass semantic object as output

36
Contd
37
Related Work

• MIT Simile Project
• Java2rdf
• Flat2rdf
• Mindswap Lab
• Excel2rdf
• csv2rdf

38
Contd

• D2R Map
• declarative, XML-based language
• to describe mappings between databases and
  ontology
• Limitations
• can create RDF, not OWL objects.
• also user expected to learn the new constructs
• user is expected to be well versed with XML

39
OWL-S Service Descriptions

• Service Profile used for discovering the
  service
• Service Model contains all functional
  properties.
• Service Grounding
• interface to plugin to WSDL description.
• Provides XSLT for transforming between XML and
  RDF/OWL.

40
Example

• Fax Service available on the Web.
• ltxslstylesheet version"1.0"
• xmlnsxsl"http//www.w3.org/1999/XSL/Transform"
  xmlnscontactOnt"http//www.flacp.fujitsulabs.co
  m/tce/ontologies/2004/03/object.owl"
  xmlnsrdfs"http//www.w3.org/2000/01/rdf-schema"
  contactOntdummy""gt
• ltxsltemplate match"/"gt
• ltxslvariable name"X1" select"//contactOntnumb
  er" /gt
• ltobjgtltxslvalue-of select"X1"/gtlt/objgt
• lt/xsltemplategt
• lt/xslstylesheetgt

41
Bio-STEER- Application of Task Computing

• Bio-informatics
• Databases and sequence matching
• Find similar proteins
• Phylogenetic analysis
• sampling properties of sequence data
• Comparative genomics
• genome content
• genome structure
• genome evolution

42
Motivation

• Problems in Bioinformatics
• Lots of tools I/O
• Lots of data structure, formats
• Lots of interaction b/w data and tools workflow
• Computationally expensive
• Wide range of skills for Biologists

43
Motivation contd

• Technologies
• Semantic Web
• Interactive composition of semantic web services
  (workflows)
• Search and information extraction
• Grid Computing
• Computational power
• TCE
• Web Services Composition
• Usability Issues

44
Services for Bio-STEER

• We have semanticized some services available on
  the web. For e.g. BLAST service, Clustal-W.
• Other services like Muscle, Phyml are being
  semanticized.

45
Demo
Map to ontology
Semantic Instances
DB
DB
DB
Access DB through WS
46
Future Work

• Mapping Object Properties directly to multivalued
  attributes.
• Link to RDF data stores like the Kowari Metastore
  to assert the instances.

47
Conclusion

• Task Computing environments are now able to
  access legacy databases and create semantic
  instances from the same.
• Extremely important in the evolution of task
  computing applications like Bio Grid Task
  Computing.

48

• Thank You

49
Questions ?