User:Phlox/Ontology notes

From Familypedia

This is for ontologies and related query/ concept information

Contents 1 Formalisms 1.1 IS_A / AKO links 1.2 Is part of (aggregation) links 2 Modeling uncertainty 3 Software support 3.1 Queries

[edit] Formalisms

[edit] IS_A / AKO links

Category is used for IS_A hierarchies, because queries can fetch the subcats. EG: to constrain a search to people who resided in a geographic area, they could select from any of the cats to narrow or broaden the scope of the search. For instance a search on residents of Gloustershire would pick up all people formerly or currently residing in any of the cities or communities in the subcats (however deep) of that category.

• Residents of United Kingdom
  • Residents of England
    • Residents of Gloustershire
      • Residents of Bristol
        • Residents of Staple Hill
          • Residents of 8 Seymour Road (for large residential buildings and complexes, people might cat them.)

(real world example: Florence Eveline Jenner (1901-1994) who lived here in 1950.

For employment, this could be useful. EG, consider a constraining a search to people that worked for the US federal government. The person could be encoded as having worked for some obscure office that a researcher didn't even know was US federal government operated. EG, the subcat structure might include:

• Worked for USGOV
  • Worked for USDOI (United States Department of the Interior)
    • Worked for USGS (United States Geological Survey)
      • Worked for Patuxent Wildlife Research Center

[edit] Is part of (aggregation) links

If queries can access nested cats, are the good for aggregations? Consider a cat structure that modeled an IS_PART_OF hierarchy:

• Family members
  • Spouse
  • Children

Note this category doesn't help us. Everyone at some point is a member of a family, but which family? This cat structure is synonymous with category person.

[edit] Modeling uncertainty

• Problem- we know that Abraham Lisenby (1819-1908) was born in Kentucky or Tennessee.
  • Normally we would assume state to be a single item. Make state a list, so that query on Kentucky or Tennessee will match this individual?

[edit] Software support

[edit] Queries

• does not support search in strings for obvious performance reasons. (no search for ~*County)
  • workaround: word break the terms, making a list, so that search for Greene will match Greene County and Greene Cty?