Toward a Structuralist R D F Schema
June 7th, 2009 Rafael AlvaradoWhen I get the time, I’m going to write a vocabulary creation language to support structuralist text interpretation. It will consist of two specs: one to handle the marking up the surface features of text, such as rhetorical figures and tropes. This will be based on my work with the Princeton Charrette Project and it will likely incorporate some ideas from Steven Bird’s work on annotation graphs. The second will be either an extension of or a variant of SKOS and|or OWL designed to represent extracted symbolic structures. It will incorporate predicates to handle relations of signification, such as has_part, has_analogy, and has_metonym, between the elements represented in the first language. At a larger level, I want to represent holistic dimensions such as context and level, as well as narratological things like encompassment, transformation, inversion, and liminality.
One of the big problems I see in this project is an apparent limitation in RDF to support triples about triples. For example, an analogy is a relation between structures, not terms. The assertion A : B :: C : D is, at minumum, an assertation about the relationship between two assertations, A : B and C : D. (The predicate of the assertions themselves is usually X has_part Y.) An anology looks something like this then:
[A has_part B] just_as [C has_part D]
The easiest way to accomplish this task would be to provide URIs for each RDF triple. I haven’t seen a general solution to this problem. I know I can create local URIs within a specific triple store, and use these in triples. But I need to define an RDF triple as a datatype first. And I anticipate problems further downstream; I wonder if the current RDF toolset is designed to handle indexing and inferencing of these kinds of triples.
If anyone has suggestions about how to handle this issue, I’d be glad to hear them.
AFTERTHOUGHT:
After writing this, it strikes me that to say that two triples are analogous is just to say that they share a predicate–so long as that predicate is sufficiently specified. To assert an analogy, then, is to assert that such an identity is important or relevant in a certain context.
