Data-driven Ontology Engineering Framework

1
Data-driven Ontology Engineering Framework

• Steve Leung, Oscar Lin, Dunwei Wen
• ICCE, 2008, Taipei

2
The Problem

• Need to find a machine readable job objectives
  for courses planning.
• Courses planning depends on a complex and dynamic
  factors.
• Predetermined categories of job objectives do not
  work
• Perceived meanings
• More than one category

3
The idea
4
The design
5
Advantages

• Data-driven
• Limited human intervention
• Human intervention does required!
• Automation
• No need to integrate new and existing ontologies
• Self correction and adaptation

6
Assumptions and Limitations

• Unknown universe of terms
• Light-weight ontology
• Application oriented
• Rigorous sampling method required
• Human expertise
• Clustering method and distance functions
• Majority threshold
• Consistency maintenance and versioning

7
Interim results

• E-advisor
• Survey
• Population all MSc IS students
• Questionnaires two open-ended questions
• 135 valid respondents
• 10 clusters
• Two concepts

8
Future works

• Quality guidelines
• Users evaluation
• Versioning system