While LITMUS is in development, I wanted to take time and answer the top three questions I get asked by researchers, academics, and other interested people I meet in sessions (among other informal settings).
1. What will your ontology–and linked data in general–actually allow us to do?
Originally taken from philosophy and now applied within the computing world, an ontology is essentially a toolkit that can be used to describe some area of knowledge on the web. In the case of LITMUS, we’re describing the realm of Irish traditional music (instrumental and song) and dance. Ontologies provide guidance on how to describe objects in linked data, and are referenced using exact web addresses for people/places/things/ideas. An example is the web-based location for the relationship member of within the Music Ontology: purl.org/ontology/mo/member_of
The ultimate goal of linked data is to provide the means to describe digital objects on the web using structured data that is machine readable, making connections between objects at a granular level. There are three components to each statement of linked data: subject, predicate, and object. The subject can be something like a person/place/thing/idea, the predicate is a relationship of some kind, and the object can also be a person/place/thing/idea.
An example of this triple (three component description) is the relationship between John Kelly Sr., the renowned fiddle and concertina player from Clare, and his son, fiddler James Kelly. We can say in regular, natural language that James Kelly’s music is/was influenced by the music of his father John Kelly Sr. In structured data, however, we need to specify exactly the person we mean (separate from other James and John Kellys in the world) and reference the relationship type from an already-published ontology. All of these references are web addresses that lead to more information or a record of the person/place/thing/idea:
<https://www.wikidata.org/wiki/Q3161199> (James Kelly’s Wikidata ID)
<https://musicbrainz.org/artist/fcca6c59-415b-4246-bd9d-dde4fbc55c97> (John Kelly’s MusicBrainz ID)
*This URI is no longer working, however it is still referenced at vocab.org.
One exciting part about LITMUS is that it will provide even more detailed relationships to describe Irish traditional music and dance objects than exist within current ontologies. This is especially helpful when we want to mirror the language that musicians and dancers use to talk about what they do.
In developing these relationships, samples taken from 50 years of album notes for traditional music have given relationships such as “playedRegularlyIn” or “firstHeardSung.” With both of those relationship examples, interesting links between music, people, and places can be made and explored when used to describe actual objects. Questions can then be posed and answered. For instance: How many musicians have said they first heard x song from a particular singer? Which tunes have been described as being “playedRegularlyIn” County Limerick?
A feature of linked data description is that multiple relationships can be applied to the same sets of objects or between two people/places/things/ideas. The same two people can be said to be “associatedWith” and “influencedBy” one another, along with many other possibilities.
When the ontology is complete, it will be published to the web and then it can be used for description of digital objects. Because LITMUS focuses on the development of the ontology, it will not progress to the point of large-scale implementation (converting ITMA’s entire catalogue to linked data, for example), but will include a use case applying the ontology to a selection of digital objects from ITMA’s collection.
2. What kinds of sources are you using to develop the ontology?
I am developing properties (relationships) for the ontology using a sample of over 30 commercial album notes taken from over 50 years of recordings, spanning from 1962-2015. The content of these album notes imitates situations in which musicians describe music in person, such as when announcing sets/songs from the stage, in a session, or in conversation with students. The sample includes solo and band recordings–both instrumental and vocal (Irish and English)–with male and female musicians.
This is an example of text and relationships derived from album notes. The album is One Out of the Fort from fiddle player Johnny Henry, released in 2012 using recorded material from 1964, 1973, 1977-8, and 1981. The notes were written by fiddler James Kelly.
Sources for other aspects of the ontology include published academic texts that give insight into the structure or organisation of music and dance as well as terminology, includng: The Companion to Irish Traditional Music, 2nd edition (Vallely, 2011); Narrative Singing in Ireland: Lays Ballads, Come-all-yes and Other Songs (Shields, 1993); and, A Selection of Irish Traditional Step Dances (Tubridy, 1998). Other sources include the article “Ireland” in Grove Music Online written by Harry White and Director Emeritus of ITMA Nicholas Carolan, particularly the sections on “Instruments” and “Traditional Music.” ITMA staff have proved invaluable resources when looking to determine structure and context of traditional music and dance practise.
3. How will you determine particular relationships, such as when something is “influenced by” or “associated with” something else?
Often we know, either explicitly or implicitly, that these relationships are there. When examining album notes or published resources, relationships are largely explicit or present in the text. We cannot always rely upon relationships to be documented through published sources, such as within album notes, memoirs, or academic books. Those relationships that are not found within published sources and that instead are grounded in oral tradition can still be discovered in a number of ways.
First, there is usually one or more persons who can act as reliable sources of information, such as family members, students, or friends. We might have materials housed in ITMA’s collections, such as audio recordings or videos, in which musicians or dancers are heard describing and/or narrating the content. This description/narration might not be captured in any additional kind of transcript or text, but is there for reference and could be used to describe the item in linked data.
We can also make careful inferences regarding relationships. If we know a musician got a tune or multiple tunes “fromPlayingOf” Willie Clancy, and this musician also “playedRegularlyWith” Willie Clancy, we might also infer that this musician was “associatedWith” him and, perhaps–if enough other relationship evidence supports it–infer s/he was “influencedBy” Willie Clancy.
What makes linked data powerful is its descriptive detail and nuance. These same strengths are what makes implementing it using cultural heritage materials all the more challenging!