The goal of this project is to create a database of the early Irish Genealogies. To achieve this goal and due to the nature of the source material, the curators chose the Resource Description Framework (RDF) to represent it. Because this is a human curated database, a human readable representation of RDF was needed, which, in this case, the curators chose the TRiG concrete representation of RDF. It is recommended for those who may not have experience with RDF serializations to read the Turtle specification first before reading the TRiG one.

The database is structured by dividing the genealogies by manuscript. Each manuscript is given its own directory which is derived from its common scholarly abbreviation. For instance, all genealogies that are derived from the Book of Leinster are placed in the LL directory. As for the ontologies, these are placed in the top level directory.

Each genealogy is divided into its "items" which represent one Turtle file in its directory. The item file name is created from its manuscript header. For instance "Aisneidem Di Araill" from the Book of Leinster has the file name aisneidem_di_araill.trig.

The curators have not always been consistant in the naming of the items. Especally in LL, "Genelogia" or "De Genelach" have been omitted.

Within each item, the individual entries are given a URL to represent that particular entry in the genealogy. The URL for an individual entry, which constitutes a node in the RDF graph, is generated from the instance of their name directly from the manuscript. If the same name appears in exactly the same form appears, whether or not it is the same person, then the first eight characters of a UUID generated by uuidgen -r are appended to differentiate between the different instances is added. For example,

<#CindFhaelad>
    a foaf:Person;
    irishRel:genName "Cind Fhaelad";
    irishRel:nomName "Cenn Faelad";
    rel:childOf <#Airnelaig>.

<#CindFhaelad-6e827350>
    a foaf:Person;
    irishRel:genName "Cind Fhaelad";
    irishRel:nomName "Cenn Faelad";
    rel:childOf <#Gairb>.

At the present moment, all URLs are prefixed with http://example.com because a permanent URL has not been purchased at this time. For example, a full URL for <#CindFhaelad-6e827350> would be http://example.com/LL/ceniuil_lugdach.trig#CindFhaelad-6e827350.

Each item belongs to a manuscript and while this is represented in the URL as described above, it is inconvenient to address the manuscript itself. To allow for this and to allow queries which are easily narrowed by manuscript, an extention to the triple format, called a TRiG, is used. This extention allows for the use of Named Graphs (see more here). In the case of this project, the manuscript is identified by its URL and is the named graph for the triples. For instance, from aisneidem_di_araill.trig:

<http://example.com/LL> {
<>
        a dctype:Dataset;
        dcterms:title "Aisneidem Di Araill"@sga;
        dcterms:isFormatOf <http://www.ucc.ie/celt/published/G800011F/text028.html>;
        dcterms:format "application/trig" ;
        prov:asDerivedFrom <http://www.ucc.ie/celt/published/G800011F/text028.html> .

     <#Conchobuir>
        a foaf:Person;
        irishRel:genName "Conchobuir";
        irishRel:nomName "Conchobar";
        rel:childOf <#Fhactnai>.

     <#Fhactnai>
        a foaf:Person;
        irishRel:nomName "Fhactnai".
}

This snippet identifies these triples as being a part of the <http://example.com/LL> graph. In this way, queries can be done on particular graphs and the user can programmatically determine which triples belong to which manuscript.

While each entry in the genealogy has its own URL, many references are to the same individuals. To represent this, owl:sameAs is used to link these URLs together. This is done: within a single item file, across item files in the same manuscript, and across manuscripts. This ensures that the various versions of the genealogies are referenced together.

Occasionally, individuals will have alternate genealogies. For ease of curation, these alternate genealogies are attached directly where they appear in the manuscript. This will often make an individual look like they have three or more parents.

There are many instances where there are individuals who are mentioned but have no name. RDF blank nodes are used to identify the individual. The curators chose a format which uses a _:missing plus a UUID fragment like above. For instance,

_:missing-04015614
    a foaf:Person ;
    foaf:gender "female" ;
    rel:parentOf <#Conmáel>, <#h-Ér>, <#Orbba>, <#Ferón>, <#Fergna>;
    rel:childOf <#Militis>;
    rel:siblingOf <#Díl>.

The alternate form of the blank node is used where convenient. For instance,

<#FiachachLabrinne>
    a foaf:Person ;
    irishRel:genName "Fiachach Labrinne" ;
    irishRel:nomName "Fiachu Labrainne" ;
    rel:spouseOf [
        a foaf:Person ;
        foaf:gender "female" ;
        rel:childOf <#MugáethMórólach>;
        rel:parentOf <#ÓengusaÓlmugáetha>
    ] ;
    owl:sameAs <http://example.com/LL/senchas_síl_ébir.trig#FhiachachLabrainne>.

Often important individuals are credited with founding a clan or tribe. In this case the population group is created as its own URL which is constructed using the same principles as for a person, as above. For instance:

<#Coscrach>
    a foaf:Person;
    irishRel:nomName "Coscrach";
    rel:childOf <#Lorcan>;
    irishRel:numChild 12 ;
    irishRel:ancestorOfGroup <#ClandCosraig>.

<#ClandCosraig>
    a irishRel:PopulationGroup ;
    irishRel:PopulationGroup "Cland Cosraig" .

Occationally, in the manuscript sources, there is more information about an individual which is added to the entry by using rdfs:comment. This is done because the curators wished to capture relevant non-structured information to capture the context of an entry. For instance,

<#Lachtna-32e54830>
    a foaf:Person;
    irishRel:nomName "Lachtna";
    rel:childOf <#Cennétig>;
    irishRel:numChild 0;
    rdfs:comment "is é ro gab ríge dar éis Cennetig. Unde dicitur Grianan Lactnai i Creicc Léith...".

There is a supplemental database named Irish Narr which contains some geneaological information which is extracted from early Irish narrative sources. This is sometimes useful for explicating some of the information in the genealogies and thus can be added to a triplestore to enhance IrishGen with narrative source information.

There are several utility Perl and Python scripts which ease the creation and curation of the database. Look in the utils directory for more information.

More specific information about the project can be found on the blog: IrishGen Occasional Topics.