Much of the metadata I’m used to working with functions as descriptions for “document-like objects” such as books, photographs, manuscripts, etc. While there is work to be done on figuring out how to represent that data in LOD-friendly ways, I’m also very interested in exploring how to represent the complex fabric of history by modeling the people and objects that may (or may not) be related to those resources. For every ship launched by Wilmington’s shipyards were hundreds of men and women who made it possible. Some of them were the captains of their own companies, other were the rank and file of shipwrights, founders, boilermakers, welders, naval architects, plumbers, electricians and scores of other professions. Owners commissioned ships, captains led them into battle, across the lines of blockades or into the teeth of a storm. Each of them may have left their mark in different ways, their footprints left on plans and drawings, in published specifications, or in company newsletters.
Representing people in linked data seems to have mostly fallen to the Friend of a Friend (FOAF) specifications, although this seems better suited as a digital business card for a contemporary person. Let’s look at some of the alternatives for providing a more robust set of properties for historic individuals. I haven’t been able to find any of the Wilmington shipbuilders represented, so I’ll use some examples of other people (surprised about how thin the representation is of American shipbuilders in Wikipedia). Like many of these entries, my own information on the individual people varies. I don’t know how many biographical essays I want to write, but I would at least like to create a record of vital statistics.
- Donald McKay
- John Ericsson
- John Roach
I was able to find some classes for shipwrights via a dbPedia record for Donald McKay. But I find the lists of many different properties from different ontologies a little confusing. Its unclear from this presentation how they all fit together or where you might find more documentation of these classes. For this reason, I often find Freebase a better place to start, because the interface is a little more user-friendly.
Here are some of the schemas that Freebase has that seem relevant to some of the people I want to talk about:
Encoded Archival Context – Corporate, Persons and Families (EAC-CFP)
Many of the people we are interested in may be represented in other kinds of library/museum authority data. Are there properties of interest in the Encoded Archival Description – Corporate, Persons and Families schema that are not represented in the Freebase schemas (and vice versa?)
Metadata Authority Description Schema (MADS)
Creating authority records to disambiguate different forms of names is also important within the library community. For example, here is the record for John Ericsson in the Virtual Authority File. Not quite linked data yet, but a new standard is making it’s way into daylight: the Metadata Authority Description Schema (MADS) in RDF. I haven’t dug deeply into this, but from a quick glance it seems that MADS is more about modeling authority records not the things that may the subject of authority records (e.g. people, etc.) That’s not to say this might not be helpful for constructing representations using Freebase, especially if those can point towards an authority record. But not sure this is a place to start with the problem at hand.
Just out of curiosity are there conversations going on within LOD-LAM about how EAC-CFP & MADS intersect?
Publishing Linked Data
OK, so it looks like the Freebase properties may be the best place for me to start. I think what I need next is a strategy for how I might publish the linked data I create. Do I publish it myself on my own servers? Should I start with Wikipedia? Freebase? Some other place? If I publish my own, what would I need beyond RDF, etc. I’ll start with the Freebase documentation and this article Publishing Linked Data on the Web