Ontology Quality Metrics

1
Ontology Quality Metrics

• Debdatta Chatterje
• Michael Moore
• Samir Tartir
• February, 6 2005

2
Why do we need it?

• More and more ontologies are being developed.
• No current way to automatically assess ontology
  accuracy, completeness, uniqueness, timelines and
  consistency.
• Currently, it is the responsibility of the
  ontology human expert to build an
  as-good-as-he-can ontology.
• Some applications may need some features to exist
  in ontologies they work on.

3
Goals and expected outcomes

• Define a set of metrics to be used to measure
  different features of ontologies and their
  instances.
• Build an application that uses these metrics.
• Use the application to compare different
  ontologies (e.g. SWETO vs. TAP)

4
How do we measure it?

• Three approaches
• Schema-based
• Instance-based
• Combination-based

5
Schema-based metrics

• Richness How diverse relations are (are they
  mostly class-subclass?).
• Width measuring out-degrees.
• Depth (Specificity).

6
Instance-based

• Connectivity how instances are distributed in
  the knowledgebase (e.g. are there any islands?)

7
Combination

• Class Completeness Can be used after adding an
  attribute to specify the expected number of
  instances per class, to represent how full a
  class is. Can also be used to see how applicable
  the ontology was to the domain it was intended
  for.

8
References

• Vipul Kashyap Trust and Quality for Information
  Integration The Data-Metadata-Ontology Continuum
• Nicola Guarino Foundational Ontologies for the
  Semantic Web
• Nicola Guarino The role of ontologies for the
  Semantic Web (and beyond)
• J Llinas, E Little An Ontology Action Plan for
  the Information Fusion Community
• Arash Shaban-Nejad Ontology (Tutorial III)
• Nicola Guarino WP3 Content Standardization
• K Supekar, C Patel, Y Lee Characterizing Quality
  of Knowledge on Semantic Web.