Design Strategy for Knowledge Base Formation to Automate a Course Map Creation

1
Design Strategy for Knowledge Base Formation to
Automate a Course Map Creation
Susan Lukose svlukose_at_olemiss.edu Susan Lukose svlukose_at_olemiss.edu



2
Presentation Outline

• Background
• Issues
• Motivation
• Existing Methods
• Our Approach
• Evaluation

3
Background

• Formal Learning - traditional classroom learning
• Non Formal Learning - online learning
• The Institute of Advanced Education in Geospatial
  Sciences ( IAEGS )
• NASA funded project
• Organization and structure of courses
• Virtual Portal Application, Knowledge Delivery
  Engine and various supporting tools
• 25 online courses

4
Background

• Procedure to create a course involves
• Domain experts
• Selection of concepts and their structure to form
  the course and creation of the content material
• Course map (Detailed outline of a course)
• Peer review by other domain experts
• Create course content which includes text,
  images, videos, question and exercises
• Professional animators create these animations
• Peer reviewed to ensure desired quality of the
  material

5
Snap-Short of a Course Map

6
Issues

• Requires extensive domain specific knowledge
• Expensive in terms time and money
• Casey and McAlpine (Casey et al. 2002), state
  that,
• Anyone who has had to create learning
  materials from scratch knows just how labor
  intensive and time consuming the process can be
• 1 to 2 months for course map creation
• 6 to 8 months for entire course creation
• No direct interaction of students and teachers

7
Motivation

• Reduce course creation time
• Option of customizing the course

8
Existing Methods

• No existing research to reduce course creation
  time.
• Digital libraries are available
• How to reuse these resources for course creation?
  
• Customizing courses
• Dicheva et al.2004, Cardinaels et al. 2005,
  Kotzinos et al. 2005, Tane et al. 2004, Mittal et
  al. 2006
• Create alternate learning path
• But no automatic method to create alternate paths
• Ontology approach used in Semantic Web
• Li et al. 2000, Cimiano et al. 2005, Wang et al.
  2006, Bisson et al. 2000, Cimiano et al. 2005
• Ontology is created either manually or from
  document corpus

9
Approach

• Use ontology (hierarchical structure of concepts
  and their relationships)
• Ontology is created from glossaries instead of
  documents

10
Approach
Glossary Course Documents
Concept Identifier Ranker

• Knowledge Base
• Glossary
• Domain Documents

Concepts Weights

• Weighted Ontology

Concept Relationship identifier Ranker

1. Super-ordinate
2. Sub-ordinate
3. Associative
4. Equivalence
5. Co-occurrence

Super-ordinate, Sub-ordinate, Equivalence
Relationships Weights
Associative, Co-occurrence Relationships Weights
Weighted Ontology

1. Document Frequency
2. Glossary Frequency
3. Hit ratio

Weighted Directed Graph
Weighted Undirected Graph
Transitive Reducer
Acyclic Graph Generator
11
Evaluation

• A case study was conducted with IAEGS
• Photogrammetry and Remote Sensing domain was
  used
• Glossaries used are
• ASPRS
• CCRS
• LDEO
• Ontology created consist of
• 11,500 concepts
• 18,500 relations
• Forest and is very sparse
• Evaluated using precision and recall measure

12
Thank You