I now have a new improved version of my family tree up as linked data here. To produce this family tree I converted the original family tree that my parents created using a perl script that takes GEDCOM to RDF. I then manually cleaned up the RDF to get the URIs in a form that I wanted.
This resulted in an RDF file giving information about parent/child, sibling and spouse relations for my family members. The vocabularies (or ontologies) used for this were FOAF, BIO and RELATIONSHIP.
I was interested in displaying more than just parent/child, sibling and spouse relationships and decided a simple extension could be to have grandparent/grandchild and ancestor/descendant information. To compute this information I used the Protege 4 OWL 2 editor. To compute grandparent information I used a property of OWL 2 called “property chains“. The property chain for computing grandchild relationships from child ones was straightforward:
childOf o childOf -> grandChildOf
(or for those who prefer rules: childOf(?x,?y) , childOf(?y,?z) -> grandChildOf(?x,?z) )
This simply states that “the child of the child of someone is a grandchild of that someone”.
The ancester information was event more straightforward to compute. Here we just make the property parentOf a subproperty of ancesterOf and then make ancesterOf a transitive property.
Given the two axioms above we can then let the OWL reasoner in Protege 4 do all the hard work and compute the implicit relationships based on the explicitly stated ones. Anyone interested in using OWL to compute more family relations should read this paper by Robert Stevens.
So I now have some RDF containing parent/child, ancester/descendant, sibling and spouse relationships. Also in this data are notions of family groups and information about birth  and death events. These events contain information about dates and places (given as text) of birth/death. Having this information as literals is not very interesting as it means I then have to go and use Google (or similar) to find additional information about the dates/places. To get round this (and create some links in my linked data) I decided to connect the places of birth/death to the corresponding resource in DBpedia (an RDF version of wikipedia) and do similarly for the dates . An example of this can be seen here http://www.johngoodwin.me.uk/family/event1917. This means I can now find additional information about a persons place of death/birth by following the links in the data if I should choose to do so. To link birth/death events to dates/place I used the event ontology.
In order to host the data as linked data I used the Talis Platform and the Paget (2) PHP library.
There is a SPARQL endpoint for the data here. We can use this to query for my uncles as follows:
PREFIX rel: <http://purl.org/vocab/relationship/>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX family: <http://www.johngoodwin.me.uk/family/>
select ?uncle ?name
family:I0243 rel:childOf ?parent . ( finds my parents)
?parent rel:siblingOf ?uncle . (finds my parents siblings)
?uncle foaf:gender “male” . ( find the male siblings)
?uncle foaf:name ?name . (this returns their names)
My next plan is to build some mash-ups using this data. Such a mash-up could use resources on the web of linked data to find famous people born in the same place/year as various family members, identify BBC programmes that are about said places etc. etc.
Now all I need to do is find a long lost relative who is also into genealogy and linked data so I can connect some nodes…what are the chances???
 – for obvious privacy reasons no birth information is given for people still living.
 – this was a fairly tedious manualish process – but some scripting helped.