Using Schema Matching to Simplify Heterogeneous Data Translation

1
Using Schema Matching to Simplify Heterogeneous
Data Translation

• Tova Milo, Sagit Zohar
• Tel Aviv University

2
Introduction

• There are large amounts of data available on the
  Web but the format of the data is not
  homogeneous.
• Most applications can handle only one or a small
  number of formats.
• There is a need to translate data from one format
  to another.

3
Introduction

• Two approaches to translating data
• A specific program to translate from format A to
  format B. (e.g. Latex to HTML)
• Data translation languages.

4
Introduction

• The solution TranScm
• A data translation system
• Automatically translates a portion (often a large
  portion) of the desired data
• Does not replace data translation languages, but
  reduces the amount of programming needed in them

5
TranScm Architecture
Input Schema
Output Schema
Import/Export Library
GUI
Rule Base
Matching Module
Typing Module
6
Data Model

• Tree (Forest) Model
• Similar to OEM
• Allows an order on children
• Can handle cyclic structures using ids as
  pointers

7
Data Model
Article
title
authors
sections
author
author
Conceptual Concepts
Al Gore Ithm
G WWW Bush
8
Schema Model

• Labeled graphs
• Some nodes may be ordered
• Each vertex is a schema element (type)
• Labels carry information about the node

9
Schema Model
Article 3
author 1
string
sections 2
title 1
authors 0,,-gt
ref
string
10
Rules

• Rules are the basis of the matching and
  translation
• Rules have an associated priority

11
Rules

• Each rule has two components
• Matching component
• Match function
• Decendents (sic) function
• Translation component
• Translation function

12
Matching

• The Match function examines schema labels to
  determine possible matches.
• The Decendents function checks the numbers and
  types of the children of the current node.

13
Matching
Article
Article
authors
author
author
author
author
14
When Matching Fails

• Matching can fail for two reasons
• Something in the source cant be matched to
  something in the target with the current set of
  rules.
• Something in the source matches several items in
  the target equally well.

15
When Matching Fails

• Via the GUI, the user can do the following
• Add
• Disable
• Modify
• Override

16
Translation

• Using the mapping generated from the Matching
  step and the appropriate rules, data is
  transformed from the input schema to the output
  schema.
• The translation process can make use of data
  translation languages
• The translation process can perform type checking.

17
Conclusion

• TranScm
• Provides a general mechanism for data translation
• Handles the common relatively simple translations
  automatically
• Can use data translation languages for more
  difficult translations