Library of Congress Subject Headings for Ontology - a review
Now that the Library of Congress Subject Headings are back online as linked data - on id.loc.gov rather than lcsh.info - we can start doing some cool things with them. They are online as XHTML+RDFa, and as RDF/XML, N-Triples and RDF/JSON. They use the SKOS vocabulary to describe relationships between concepts, and you can even download the whole dataset in RDF/XML or N-Triples (merge the whole thing into one file and you’ve got a nice performance test if you happen to be one of the unlucky few to be responsible for an RDF processing library (I’ll get back on that once my academic committments are out of the way for the year).
Anyhow, as part of an idea I’m working on to try and tie together my interest in the Semantic Web and my interests in academic philosophy, I started poking around at the subject headings in philosophy. The first thing I did was to search for ontology in the subject headings. Interesting: not just Ontology, but history of ontology and ontology in popular culture! Okay, in literature and motion pictures. Still awesome. But I must focus. So, I go to ontology. I want to try and find some headings which are applicable to a book I have spent a fair amount of time reading this year, and recently bought a copy of: Universals: An Opinionated Introduction by D.M. Armstrong. The name gives away the subject: universals. Universals is the name of one particular answer to the problem of universals. The problem of universals may also be answered by suggesting that universals don’t exist but something else does. It’s a bit of an amiguous term. There are books which discuss the problem of universals but which don’t talk about universals as a solution in much depth (or at all). There are broadly three solutions to the problem: one is realism, where properties are universals; one is nominalism, which gets rid of the need for properties and therefore doesn’t need to explain anything; finally there is what’s now called trope theory (but has been called particularism), which explains properties as tropes rather than universals. As this isn’t an essay on tropes but on categorisation of books, I won’t go any further into what these solutions are. If you want to know, I’d suggest reading the aforementioned book by Professor Armstrong.
Now, my university library classifies the book under three categories: Universals (Philosophy), Nominalism and Realism. Now, the first two are broadly useful categorisations. My university classify into Nominalism a book titled Pictorial nominalism on Marcel Duchamp’s passage from painting to the readymade, but it’s broadly right. The realism category covers a large amount of different types of realism: realism about universals, moral realism, scientific realism and the critical realism of aesthetics. But there’s more. The book also talks about trope theory - as any comprehensive book on the subject ought to. There is a category on tropes. There are also categories on relations and substance, both of which are covered by the book in some detail.
These categories are not tied together in any reasonable hierarchy. That’s what’s missing. Tropes, nominalism and universals are all a branch on the SKOS tree - but they aren’t represented thusly.
But it gets more interesting than that! The book discusses a number of philosophers - not exactly difficult to imagine. It discusses John Locke, Bertrand Russell, C.B. Martin, Anthony Quinton, J. Cook Wilson and others. I’m guessing they will have representations on the Linked Data web soon enough.
There’s a dbPedia strangeness I’ve just seen in the John Locke entry. Locke apparently has the dbpedia-owl:influenced property towards Arthur Schopenhauer and Aristotle. Whichever direction this property represents, it cannot be so! Aristotle may have influenced John Locke, and John Locke may have influenced Schopenhauer - but Schopenhauer could not have influenced John Locke, nor could John Locke have influenced Aristotle. Some inference of this kind ought to be possible from the birth date and death date (the latter of which does not exist for Locke). I should probably add this as a test as part of the new dbpedia-tests I’ve started.
Despite all my complaints, I am really quite excited by all the linked data on the web now: dbpedia, LCSH and much more. Very awesome, and great potential to do interesting stuff.
