Using Schema Matching to Simplify Hetereogeneous Data Translation

1
Using Schema Matching to Simplify Hetereogeneous
Data Translation

• Tova Milo
• Sagit Zohar
• Presentation Eddie Campbell

2
Why data translation is important

• A large variety of data exists on the web
• This data is of heterogeneous sources
• As demand for the data increases so does the need
  for efficient translation
• Schema Matching can simplify translation

3

• Examples of data formats
• Specific vendor format database
• SGML
• LaTeX
• DX
• CAD / CAM
• Etc
• Companies use applications that can only handle
  some data formats.

4
Naïve Approach

• LaTeX -gt HTML
• HTML -gt .txt
• Specific dedicated translators. From personal
  experience
• HTML -gt .txt -gt relational db bad idea!
• The need for a general framework exists

5
TranScm

• A rule-based system data translation system.
• Observation 'in many translations the schema of
  the target system is closely related to that of
  the source data both schemas aim to represent
  the same data.'
• Each rule
• Defines a possible match between schema
  components
• Provides means for translating instances
• Rules may be supplemented during translation by
  user input (programming).

6
TranScm Matching Process

• Not fool-proof
• 1. A component from source can not be matched and
  it is not known if it is redundant
• 2. A source component matches gt1 target component
• Solutions
• 1. User adds rules to cope
• 2. User specifies best match
• Spell-Checker analogy?
• TranScm simplifies data translation not a
  replacement for previous methods!

7
DTD
OODB Schema
8
Middleware Schema

• A common framework data model is required.
• Data providers wishing to participate must supply
  mappings to / from schema
• Schema represented by Graphs
• Data Instances by Labeled, partially ordered
  forests

9
Schema Graph in middleware
10
(No Transcript)
11
Explicit form of data instance in Middleware,
this form exists then data instance Conforms to
graph.
Complicated graphs can be flattened
12
TranScm Architecture

• Rule base
• Matching Module
• Typing Module
• Gui

13
Conclusion

• TranScm is an example of a data translation
  system that uses schema matching
• Schema Matching techniques are used to simplify
  data translation not replace existing techniques
• User dependent!