Steve Hays & Ian Johnson. Building integrated databases for the web

Archaeological Computing Laboratory, University of Sydney

Novel data modelling approach. Real world relationships aren't as simple as what is modelled by entity relationship diagrams; there can be multiple contingent relations changing over time, and entities can split into complex types as knowledge grows.

Heurist knowledge management model: start with table of record types, then table of detail types (= fields), and requirements table, binding details to records and how they behave. Summary data is stored in a record, detail information are stored as name/value pairs. Relationships are modelled as a first order record. (Reifying the relationship allows it to have attributes.)

Raw querying performance is poor; can't use complex SQL queries; obscure to explain. But performs acceptably with 100k records; export to RDF triple store with SPARQL to improve performance. Increase in flexibility will outweigh drawbacks.

Point is to create a meta-database, linking info to info across archives. (Would like to use persistent identifiers to do so.)

No comments: