Mapping%20Target%20Schemas%20to%20Source%20Schemas%20Using%20WordNet%20Hierarchies

1
Mapping Target Schemas toSource Schemas
UsingWordNet Hierarchies

• Masters Thesis Proposal
• David Jackman

2
Biskup-Embley Approach
Target View
Source Schema
Other Source Schemas
3
WordNet Background

• A lexical database for English
• Developed at Cognitive Science Lab at Princeton
  University
• Has become a standard in natural language
  research
• English terms are organized into synonym sets,
  each representing one underlying lexical concept

4
Framework
Target Schema (XML)
Mapping Process
Integration Mappings (XML)
Source Schema (XML)
5
Mapping Process

• Written in Java
• For this research, two procedures are used
• First creates mappings with a WordNet score
• Second adds a context score to those mappings

6
Target
Source
7
Target
Source
Target Term Distance Source Term Distance Root Term
Publication 3 Article 2 Creation
Publication 0 Journal 2 Publication
(Pub.) Name 0 Title 2 Name
(Pub.) Name 5 (Pub.) Date 4 Social Group
(Pub.) Name 1 Article 4 Language Unit
(Pub.) Year 0 (Pub.) Date 2 Year
Author 0 Author 0 Author
Abstract 0 Abstract 0 Abstract
Abstract 2 Title 1 Writing
8
Target
Source
9
Experiments

• Run in 4 domains Music CDs, Genealogy, Real
  Estate, Travel
• Source schemas taken from existing web sites
• Target schemas taken from standard XML DTDs or
  created by independent third parties
• Combined WordNet Context scores compared
  against mappings determined by a human expert

10
Conclusion

• Data integration process can become more
  automatic
• Combination with research being done in data
  extraction and data warehousing