Information Integration

1
Information Integration
Dagstuhl Workshop on Information
Integration April/May 2002

• (My Personal) Workshop Summary

2
Overview

• Supported by
• Focus Group Results (esp. Christoph Freytag,
  Rakesh Agrawal, Gerd Stumme)
• Panels (we all)
• Starting PointA Look at the Original Seminar
  Proposal
• The topics raised
• The solutions given
• Hot research topics
• Closing PointThe (forseeable) truth

3
Old Keywords
4
New Keywords
InformationIntegration
ApplicationIntegration
BusinessIntegration
Data Integration
Process Integration
5
Information Integration Topic Description

• II subsumes all technologies needed to provide
  for manipulation of information scattered over
  many data stores while supporting a single system
  image.
• The data stores to be integrated are inherently
  heterogeneous in nature, owned by different
  organizations, and distributed over the whole
  world.
• Data can be structured semi-structured, or
  unstructured.
• Data access can be based on standardized
  interfaces or via proprietary APIs.
• II is expected to become a key technology in many
  application areas like
• product data management
• business process management
• enterprise application integration
• life science
• entertainment.

6
Discussion Areas

• How to get Access to the various data stores?
  Different technologies like SQL/MED wrappers,
  J2EE connectors, EAI adapters, and Web Services
  can be used for these purposes.
• When should either of these technologies be used?
  
• Can they be unified?
• What are possible System Structures?
• Which role will database systems, application
  server, workflow systems, messaging systems,
  portal servers, etc. play?How do they relate and
  cooperate?
• Does Web Database Technology suffice?
• Can XML be used as the language for describing
  the integrated information base? How to capture
  navigational access based on hyper-linked HTML
  pages performed today in many application areas?
• How to combine search and query functionality?
• How is XML stored - sliced/diced, as whole
  document as file in file system, as whole
  document but combined with other documents in
  file system? How do you index these effectively?
  How do you combine SQL and an XML-based query
  over the same data (i.e., XML query against SQL
  data and SQL against XML)? Is a pure XML database
  the way to go or will an extended relational
  engine be the right solution?

7
Discussion Areas

• How is information described?Which information
  qualities are needed? How can qualities be
  compared, assessed, measured,? Which metadata is
  relevant (schema, ontologies,)?
• Which Federated Database Technologies can be
  used?What is a federated schema if structured
  and unstructured data are brought together?
  Which schema integration techniques, federated
  query and search technologies are applicable?
• Not Discussed
• Which Transaction Model is appropriate?Which
  transactional guarantees are needed? Which
  concurrency models, recovery models are
  applicable?
• Not Discussed

8
Some Results/Agreements ... Definitions

• Data Integration
• integration schema an image presenting all
  relevant facts as one data source
• generic functions for access and change
• Integration different from cooperation and
  interaction
• Integration properties are Impacted by
• Experimental/Exploratory vs. Production

• Exploratory
• loosely coupled
• fluid integration
• Data centric

• Production
• Function integration
• less flexible and often fixed

9
Some Results/Agreements ... Definitions

• Structured object
• ltoid, ltname, valuegtgt
• Unstructured object
• ltoid, wordgt
• ltoid, unknown/complex structuregt
• Semi-structured object
• ltoid, ltname, valuegt, wordgt
• ltname, valuegt pairs may be
• Given (e.g. author, title, etc.)
• Extracted (e.g. Date, Zipcode, etc.)
• Inferred (e.g. Topic)

10
Some Results/Agreements ... Definitions

• Metadata can be anything between natural language
  text and formalisms with formal semantics (e.g.
  ER models, (first order) logics, description
  logics, ontologies), including intermediate
  degrees of formality (e.g. XML, RDF)
• For supporting II, we need more formal models
  which allow for machine manipulation
• Ontologies are
• data about metadata (schema for metadata)
• Force people to use them!
• (at least) 2 Secrets/Rumors
• Late night tutorial for thursty people
• (late) night tete-a-tete (maybe separe)
  tutorial

11
Some Results/Agreements ... Web Services

• Web Services is a new model for using the Web
• "An interface that describes a collection of
  operations that are network accessible through
  standardized XML messaging.
• transactions initiated automatically by a
  program, not necessarily using a browser
• can be described, published, discovered, and
  invoked dynamically in a distributed computing
  environment

12
Some Results/Agreements ...

• Information Integration the database reaches out
  
• Unstructured Data Support
• OLAP, Mining, rich Search
• Federated database extensions
• Metadata management
• SQL and XML
• Pure XML
• ...
• SQL and XML and NF2-like technology
• DBMS should reflect some semantics of the
  applications

13
Some Results Confluence of Multiple Disciplines
14
... Other Results

• More focus on Process Models and Process
  Specifications
• DIFF operator
• Formal theory for process models etc.
• More focus on Semantics

• Much more focus on Semantics
• ...
• Performance

• Much more Performance

• ...
• Information systems should also project into the
  future ...

15
The (forseeable) Truth

• Heterogeneity is Fact

Integration is Fact
Make Life in Heterogeneity possible ... Easy!
Data Integration per se is not beneficial!
There has to be life beyond angular brackets!