Elections are tough to model ! - now Election data model version 2.

The first version of the data model for election results provided a useful experiment in model-driven RDF applications, but the data model itself was poor. I've had another go and produced a more complex version 2.

The main differences are:

  • Separation of the Poll from the Constituency, so that multiple polls, including by-elections can be represented.
  • Persons as first class entities
  • Adding Assembly so that bodies other than the House of Commons can be represented, for example the European parliament, the Welsh Assembly, the Scottish Parliament and even the Cabinet and the labour party executive
  • Membership of an Assembly is a period of association of a Person with an Assembly in a role

The main 2010 data has been reloaded and is browseable again. I've still to do the derived data for the overall Party results at the May election and other derived data as well as the 2005 election data so some of the queries don't yet work.

The model is looking better, but I still have problems with some of the relationship names which are overloaded, need inverse predicates I think, don't know if its worth keeping local names for each Class as well as rdfs:label (or should this be skos:prefLabel?), and I need more vocab reuse.  No time to do all this since we are soon to set off on our summer cruising.

This more complex model is harder to query and needs more derived triples to make it workable, but it looks like an improvement in fidelity to the reality of elections but with a cost of decreased usability.  It's a fine line to know where to stop, both temporally (how much to explicitly date) and conceptually (could US elections fit this model?). Still perhaps there is some convergence - I'm reminded of Piet Hein's poem

The Road to Wisdom?//Well it's plain and simple to express/to err/and err/and err again/but less/and less/and less.