Session held by ReiniUrban at Friday 11:30, P1
Links are not only page => another page, in a semantic wiki a link has a name (predicate) which describes a link. Every link describes a triple: subject (pagename) - predicate (verb, relation) - object (another page).
In this triple-store you should be able to search for complex expressions.
A SemanticWiki should do:
- Provide a syntax to define links enriched with relations and attributes.
- Search for those created relations and attributes. (The hardest part)
- Import / Export existing triple-stores, usually describing a knowledge base, in various (now often XML based) formats. RDF, RDFS, OWL, DAML, ...
Example:
San Diego
- is_a::city located_in::California with a population:=1,200,000.
This defines two relations: is_a and located_in and one attribute population. Attributes are usually used for numeric data, everything which is not necessarily a link. You should be able to calculate with attributes to search for larger cities, i.e. a search for <is_a::city and population > 1,000,000> => San Diego
More complicated searches would search for relations/attributes defined in other pages, "nested sub-selects", or "inferred" facts, facts derived from logical reasoning.
Since everybody knows that certain facts will contradict each other being written on other pages, this will lead to empty results. "The SemanticWeb frustration experience." (Examples needed)
I also discussed my problem with typed attributes as in semediawiki, types not automatically inferred from the value. This will lead to unneccessary overhead in import/export of facts for different engines. I want only strings (relations) and unit-ready numbers (parsable by /usr/bin/units).