Semantic in Information Systems: Current Issues New Trends

1
Semantic in Information SystemsCurrent Issues
- New Trends

• Kokou Yétongnon
• Professor

2
Summary of ideas

• semantic and Integration of information systems
  is a recurring problem
• (new variations on an old problem)
• Global semantic of multiple systems is hard to
  determine
• (large number of information sources with
  varying levels of semantic)

3
Summary of ideas

• Dynamic environments require new ways of dealing
  with semantic and interoperability
• Semantic must locally emerge from interactions
  among information systems
• (Negotiations, agreements etc)

4
Outline

• Introduction and Motivation
• Data Integration
• Modeling Semantic
• Information System Interoperability
• Semantic in Dynamic Environments
• Conclusion

5
The Traditional View of Information Systems

• Where is the semantic of the system

Schema

• semantic partially represented in the
• Schema
• Occurrences conform or agree with
• the schema

Instances / Occurrences
The traditional, safer and formal picture of
information systems the simple life
6
The Traditional View of Information Systems

• When do semantic issues arise?

• Schema is created by several designers
• Each handles a part of the system

Schema
Integrated Schema

• Semantic reconciliation easy
• Make all designers agree on
• Differences in terms or concepts

Schema1

Schema n
7
The Traditional View of Information Systems

• Main goal of the centralized system is to process
  company information

Schema

• Typical Applications
• Payroll
• Managing enterprise project or
• production system

Instances / Occurrences
8
Information Systems in Networked Environment

• Distributed systems and widespread resource
  sharing evolve from two major developments
• Powerful, efficient and smaller computers and
  processors
• efficient communication concepts
• Internet
• Wireless communication
• Ad hoc, mobile networks (sensors, PDA, other
  devices )

9
Information Systems in Networked Environment

• Data have also evolved in
• Complexity (image, multimedia, hypermedia)
• Volume (and storage requirement)
• Type
• So the related semantic is also evolving in
  complexity

10
Information Systems in Networked Environment

• The main IT goal is
• Not only to process enterprise information
• But also to sharing information
• Global need a system must handle information
  from a variety of sources (proprietary data,
  public information in web pages, information in
  web services)

11
Information Systems in Networked Environment

• What are the new challenges? Discovering and
  extracting relevant information

12
Information Systems in Networked Environment

• How does the traditional view evolve when we have
  multiple information systems?

• May not be present
• Multiple schema in distributed environments
• Schema May be expressed using different models
• Description levels, precision and power may be
  different
• Formal description may co-exist with natural
  language descriptions

Schema
Instances / Occurrences
13
Information Systems in Networked Environment

• How does the traditional view evolve when we have
  multiple information systems?

• Occurrences distributed over multiple sites
• Replicated or duplicated over sites
• Fragmented over multiple sites

Schema
Instances / Occurrences
14
Information Systems in Networked Environment
Schema
, ... ,
High level Cooperation or interoperation of
IS Information sharing
Schema
, ... ,
Low level Communication Network
Schema
, ... ,
15
Information Systems in Networked Environment

• If we could build a Global Semantics

Global Semantics
Query
Virtual Integrated Information System
Schema
Schema
Schema
, ... ,
, ... ,
, ... ,
15
16
Information Systems in Networked Environment
What are the main issues?
Schema
, ... ,
Schema
Semantics
, ... ,

• What it is?
• How to best represent the semantics of data?
• How to reconcile differences in semantics?
• What about missing semantics?

Schema
, ... ,
17
Data IntegrationA Higher-Level Virtual View
Independence of

• The basic Idea

source location
Query
data model, syntax
semantic variations
Mediated Schema

Semantic
Mappings
S1 S2 S3

ltcdgt lttitlegtThe best of lt/titlegt
ltartistgt Carreraslt/artistgt
ltartistgtPavarotti lt/artistgt
ltartistgtDomingo lt/artistgt ltprice,
USgt19.95lt/pricegt
SSN Name Category 123-45-6789 Charles
undergrad
SSN CID 123-45-6789 CSE444
234-56-7890 Dan grad
123-45-6789 CSE444
234-56-7890 CSE142


CID Name Quarter
CSE444 Databases fall
CSE541 Operating systems winter
18
Application Areas

• Business applications

Enterprise Databases
Ent. Integration Applications
Business analysis
Single Mediated View
Portals
Legacy Databases
Services and Applications
19
Application Areas

• Science (Bioinformatics)

Phenotype Structured
Sequenceable
Vocabulary Experiment
Gene
Entity
Microarray
ProteinNucleotide
Sequence
Experiment
OMIMSwiss-
HUGO GO
Prot
Gene-
Locus-
ClinicsEntrez
LinkGEO
Hundreds of biomedical data sources available
growing rapidly!
20
Application area

• Scientific Data Grid (Physics)
• CERNs EDMS
• PDM (Product Data Management)
• Engineering Data Management System
• MDAS Massive Data Analysis System
• San Diego Supercomputer Center 95-97, DARPA
  financed
• Manage resources in a heterogeneous distributed
  system
• Metadata and data description
• Detect available resources, storage spaces

21
Application Areas

• WEB
• E-commerce (Amazon.com, Barnes and Nobles)
• E-tickets reservation
• Online hotel Reservation

22
The Semantic WebBerners-Lee

• To allow knowledge sharing at the web scale
  (interaction between Machines or users)
• Web resources must be described by ontologies
  (precise explicit semantics)
• Need rich domain model
• Powerful standards (RDF/OWL)

23
The Semantic WebBerners-Lee

• Challenges
• Complex Semantic integration issues at the web
  level (This may be too complicated for non
  technical end user, unless fully automated)
• Lack of convincing applications at the semantic
  web level

24
The Semantic WebBerners-Lee

• Where are the real obstacles to the semantic
  integration?
• Systems
• Managing different platforms
• Query processing across multiple platforms
• Social
• Locating and capturing relevant information in
  the enterprise
• Convincing people to share data (privacy and
  performance reason
• Logic
• Schema and heterogeneity

25
Virtual EnterprisesWorkflow model

• Challenges
• Decentralized organizational structures
• Variety of Information Sources and Services

26
Virtual EnterprisesWorkflow model

• Issues operational aspects of the business
  process
• Interoperability
• Autonomy
• Openness and sharing information
• Dynamic Participation, Mergers and Acquisition
• On-the-Fly Integration

27
Workflow Management Syst.

• WfMS Languages
• WSDL basis for many inter-organizational
  workflow specification language
• Other languages ebXML, WSCL, XLANG, BPML
• Problems with languages
• Advanced but lack of common taxonomy
• Do not support different views of the workflow
• Solution
• Agreement based inter-organizational workflow

28
Agreement based Workflow Model

• Local Modeling view
• Each org creates a local/personal view of its own
  workflow
• Based on Loose interaction

View-b
Compatibility Analysis
WFLa
WFAb
Agent Negotiation
Global View
Agent Negotiation
Agent Negotiation
WFAc
WFLa
WFAb
Compatibility Analysis
Compatibility Analysis
View-c
WFAc
WFLa
WFAb
29
Semantic Modeling perspectives Ontologies

• An ontology provides a shared representation and
  understanding of data and services in a common
  domain
• Use ontology for interaction between people and
  application systems.

30
Semantic interoperability

• Ability of a system component to provide
  information sharing and inter-application
  cooperative control
• Ontologies are used as a comparaison reference
• Ontologies are forms of a-priory agreement on
  concepts, therefore their use is insufficient in
  ad hoc dynamic environment where all possible
  interpretations are anticipated

31
Semantic interoperability

• Interoperation of systems X and Y

Ontology Partial? Global? Predefined?
Mutual understanding of R and Y
Request R
Information system X
Information system Y
Response Y
32
Semantic interoperabilityMining semantic from
the instance
Missing values? Requires an ontology? Maintenance?
Extraction based on cluster Similar instances?
Extract Semantic Based on similarity Of instances
WEB pages containing Instances
33
Semantic interoperabilityMining semantic from
the instance

• How to define similar instances in this case?
• Bookstore Example (Amazon.com), consisting of
• Items (books, etc)
• Customers (millions)
• Each book is represented by a vector such that
  Vi is 1 if customer i bought a copy of the book

34
Semantic interoperabilityMining semantic from
the instance

• How to define similar instances in this case?
• Books bought by the same groups of customers?
• Use other clustering method
• We plot the vector in an n dimensional space such
  that each dimension represents a customer and
  each point defines the customers who bought the
  same book

35
Semantic in Dynamic EnvironmentP2P - Self
Organization Systems

• Example of Semantic Overlay P2P Networks (SONs).
• Peers with similar content are clustered together
  on content hierarchies (similar to ontologies)

36
Semantic Overlay Networksfor a music application
Rock
A
B
Rap
R
C
D
E
F
H
G
o
37
Classification Hierarchiesof the semantic
overlay network
Music
Music
Decade

10s
Now
90s
Rock
Jazz
Sub style
Music
Soft
New Orleans
Dance
Bop
Tone


Pop
Fusion
Warm
Sweet

Exiting
38
Generating semantic Overlay Networks
Query
Query Classifier
Concept hierarchy
SONs
Query Result
SON Definition
Data Distribution
Node Classifier
Document Classifier
New Nodes
39
Conclusion

• We have shown that
• Semantics is ever pervasive in information
  systems
• Semantic integration of information systems is a
  hard problem
• Because Schema when present never captures the
  intended meaning of information
• Because automatic resolution of differences is
  not yet satisfactory
• Because technological, social and logic obstacles
  still exist

40
Conclusion

• We have shown that
• Ontologies are increasingly used in integration
  solutions
• Ontologies based have limitations when used in ad
  hoc dynamic environments
• One modest objective of integration solutions
  should be to limit the required human effort

41
Thats All

• Thank You