Notes about ontology creation methodologies (2 papers)

Yesterday I unexpectedly read two whole papers about ontology development methodologies. They were open in tabs I don't remember opening, but presumably did so during our weekly Ontologies With A View meeting last Friday. There are still a bunch more tabs open with papers or articles about the same thing, so maybe I'll read those later..

The notes are here more or less as I scribbled them down whilst reading, and I haven't expanded with any analysis or discussion as of yet.

Notes in purple are things I intend/need to investigate further; colour-coding is just for me, really.

Jean Vincent Fonou-Dombeu & Magda Husiman (2011) Combining Ontology Development Methodologies **and Semantic Web Platforms for E-government **Domain Ontology Development. International Journal of Web & Semantic Technology (IJWesT) Vol.2, No.2, April 2011

• Start by describing ontology in a human-readable way, then turn to RDF (etc) to be machine readable.

• Says there's not sufficient practical research around existing technologies or ontology development guidelines that would allow non-experts in e-government domain to make ontologies.

• Uses framework from Uschold & King (see later in this post) to describe ontology - technique used here should be platform independent.
• Then uses UML to semi-formally represent ontology

• Uses Protege and Jena to convert to OWL and RDF

• Paper's goal is to produce guidelines for e-government developers to create semantic content AND strengthen adoption of Semantic Web technologies in governments (particularly developing countries).

• Outlines RDF, OWL, Protege and Jena (described as leading platforms; mentions other platforms: WebODE, OntoEdit, KAON1, Sesame).

• Very critical of other literature; either ontologies have been produced but no practical information given; they've been developed with proprietary platforms; or they're only conceptual and don't say how they could actually be constructed with existing technologies. other studies have not focused on a methodological approach, which means nothing is easily repeatable.

• Detailed comparative studies of methodologies in:

  • • M. Fernandez-Lopez, “Overview of Methodologies for Building Ontologies, ⁂ In Proceedings of the IJCAI-99 workshop on Ontologies and Problem-Solving Methods (KRR5), Stockholm, Sweden, 2 August, 1999.
  • H. Beck and H.S Pinto, “Overview of Approach, Methodologies, Standards, and Tools for Ontologies,⁂ Agricultural Ontology Service (UNFAO), 2003.
  • C. Calero, F. Ruiz and M. Piattini, “Ontologies for Software Engineering and Software Technology, ⁂ Calero.Ruiz.Piattini (Eds.), Springer-Verlag Berlin Heidelberg, 2006.

• Case study: Ontology for monitoring development projects in developing countries (OntoDPM)

  1. Create with Protege:

     • class heirarchies
     • slots
     • domain and range of slots
     • Based on the UML
     • Saved as OWL
  2. Then put content in RDF with Jena

*Mike Uschold and Martin King (1995) **Towards a Methodology for Building Ontologies. Workshop on Basic Ontological Issues in Knowledge Sharing, IJCAI-95.

• Steps:

  1. identify purpose

  2. build ontology

     • capture
     • coding
     • integrating existing ontologies
  3. evaluation
  4. documentation
• Purpose
  • Many ontologies are intended for reuse
  • Should survey purposes to clarify options for future projects
• Building
  • Capture
    • Identify key concepts and relationships in domain of interest
    • Produce unambiguous text definitions for these
    • Identify terms for these
    • Agree on all of the above
  • Coding
    • Explicit representation of conceptualisation in a formal language (choose a language)
  • When can capture and coding stages be merged?
  • Differences between building ontology and creating a general knowledge base (thinking about methodology will help with this)
  • Integrating (during either or both of above)
    • Work must be done in agreement between communities
    • Make explicit all assumptions underlying an ontology
• Evaluation
  • Judge against requirements specification (and/or)
  • Judge against competency questions (and/or)
  • Judge against real life
  • This paper looks at knowledge base systems, and adapts for ontologies.

• Documentation

  • Desirable to have established guidelines for documenting
  • Main barrier to effective knowledge sharing is inadequate documentation
  • ALL important assumptions should be documented

• Case Study

  • Main emphasis is on capture phase
  • Initially:
    • define ontology (Gruber)
    • identify users and usage (initially abstract, then clarify with real life)
    • choose language (Ontolingua was chosen)
    • choose method for capture - BDSM (IBM) supported by others:
    • KADS
    • IDEF5
    • OO Analysis and Design techinques
    • Gruber's principles for ontology design

  • Categorisation is fundamental to the human condition (Lakoff)

    • Not heirarchical, but:

    • GENERAL
      ^
      BASIC -> primary with respect to knowledge organisation
      v
      SPECIFIC

    • eg.
      SUPER: Animal / Furniture
      BASIC: Dog / Chair
      SUB: Retriever / Rocker
    • Certain concepts used subconsciously, rather than understood intellectually.
      • These have a more important psychological status.
    • Therefore paper uses middle-out approach to capture terms
      • (bottom-up = too much detail unnecessarily,
      • top-down = risks imprecision)
    • BASIC concepts first because:
      • most important
      • used to define non-BASIC terms
      • increase clarity, especially for non-technical use
      • backed by BSDM experience of paper author
• Scoping
  • Brainstorming
  • Consult corpora if there aren't enough domain experts to brainstorm
  • Grouping
    • structure terms into naturally arising sub-groups
    • collate synonyms
    • consider things that might refer to each other
• Meta-ontology
  • Don't commit too early, can restrict thinking. Let concepts and relationships themselves determine requirements.
  • Be consistent.
  • Use technologically neutral language ('thing' vs 'entity').
  • Start with areas where there's most overlap.
  • Work from basic terms to more abstract ones within an area.
• Producing definitions
  • Agreeing on definitions (varying degrees of problems)
  • Handling ambiguous terms
    • clarify ideas without technical terms
    • use a dictionary!
    • label definitions, eg. x1, x2
    • determine most important concept
    • choose a term, avoiding original ambigious one
  • Avoid new terms
  • Terms get in the way (peoples' preconceived ideas) - concentrate on underlying meaning and concepts.