Kurt Cagle is the managing editor of XML Today and is a ntributing editor to O’Reilly Media, DevX, and TechNewsWorld.  He is working with Diane Mueller on a general XBRL book for O’Reilly Media to be published in early 2010. Mr. Cagle runs a nsulting firm, Metaphorical Web, dealing with future technology issues, XML, distributed mputing, and the like.

The Semantic Technologies 2009 nference in San Jose ended recently, a fascinating event in what I’m ming to nsider the most bleeding edge arena of development on the Web today. It was perhaps more notable this year for the surprising number of "suits" in mparison to the crowd in recent years of PhD students shepherded by their respective professors. The Semantic Web is changing, maturing, and reaching a point where development is beginning to move beyond the laboratory and toward real world solutions.

Semantic Web and XBRL issues have previously been ably explored on this blog by Andy Greener and Ashu Bhatnagar. I wish to build on their work by mparing semantic technologies. 

The focus of XBRL would seem to be tailor made for the Semantic Web, specifically for technologies such as RDF, RDFa, and OWL. For those of you unfamiliar with the terms, RDF is short for the Resource Description Framework, a rather glorified term that, at the most basic level, provides a way to describe both resources (anything that can be cleanly represented as a data model, such as a quarterly filing, a set of documentation, or an invoice) and the relationships (or links) between those resources. To put things into historical perspective, XLink (which linkbases are built on) was effectively abandoned by the Semantic Web mmunity fairly early on because of the difficulties inherent in dereferencing given ntent, and RDF became the logical suessor.

RDF syntax originally was based in XML, and there is still an XML representation. However, other notations that were less verbose and dense also emerged, including the N3 notation and Turtle notation, which is a superset of N3. XML and Turtle notations are essentially synonymous, and translators for moving between RDF/XML and RDF/Turtle are freely available. Another notation — and one that holds nsiderably more potential for XBRL — is the RDFa format being supported in the W3C by Mark Birbeck. RDFa provides the same relational structures — such as the fact that a given text block is a date or taxable acunt total — that can be ended within normal RDF, but does so using attributes within HTML or other related XML ntent. This means that an XBRL "report" uld be written in a normal HTML markup, but with critical properties ended via attributes of elements to provide the detailed information for some external processor.

The OWL moniker is a little less obvious. It derives from the Web Ontology Language, the letters inverted in order to create a somewhat easier to remember acronym. OWL is a language for describing classes, class properties, and individual instances of such classes, which sounds somewhat similar to the XML Schema Definition Language (or XSD, as it’s usually now) in that it can be used to describe entities. The difference is that OWL can describe a nsiderably broader array of relationships, and those relationships can include more sophisticated "graphs" than XML (where a graph can be seen as the foundation of a data structure) because what’s being modeled is knowledge, rather than simply objects. This distinction means that it bemes possible to make inferences by following the relational assertions with the classes. OWL includes support for a query language called SPARQL (a recursive acronym meaning SPARQL Protol and RDF Query Language), which makes it possible to explore such assertions, with the results being passed back as XML documents.

A send, perhaps more critical, distinction is that unlike XML Schema languages, RDF and OWL assumes an open assertion model. That is to say, the model can change dynamically as new nditions emerge or as alternative needs arise. This differs from an XML model where in general the various properties are arranged in a mparatively limited hierarchical structure (more flexible than a relational database model, but nowhere near as flexible as RDF/OWL). This flexibility can me at some st, of urse. RDF/OWL queries are usually slower than the equivalent XQuery or relational database query operations, but given the use cases involved, this may not in fact be that big of a limitation.

If this sounds similar to how XBRL schemas are created, that’s because it is. An XBRL schema typically is a bundle of assertions, either property assertions such as the fact that a mpany’s total asset value is a specific value or that the asset value property is associated with the ntext of 3Q09 earnings. Similarly, that asset’s English long label can similarly be bound to the total-asset-value property. These are all triples, and the RDF/OWL ontology uld very easily enmpass this information and then produce the output in one of any number of formats as necessary.

Moreover, such assertion chains can produce fairly sophisticated automated analysis. For example, through sets of OWL assertions you uld get a listing of mpanies and their earnings derived from capital  investments in the 3Q09 where the reports are for mpanies involved in the airline industry, that have P/E ratios exceeding 25, and that are located in the Pacific Northwest. Such a query would be hideously mplex in SQL, doable but still mplex for an XML database, and relatively easy within an RDF/OWL database.

What’s more, an RDF triple store (which is what such an assertion database is called) can also work with information outside the spe of the XBRL in a principle called Linked Data. For instance, a send triple store might ntain a list of biographies of C-level managers for Fortune 1000 mpanies, including previous mpanies that they worked for. Because such Semantic databases employ a mmon underlying transport language (RDF), this means that such SPARQL queries uld inrporate both sets of data in order to "pivot" on a given individual in order to see his track rerd of earnings by mpany he worked for to rrelate whether he’s generally proved an asset or liability to the mpany.

The advantages inherent in an RDF representation of XBRL should make a mpelling case for being able to express XBRL documents in this manner, especially if such information uld be extracted from annual reports or other reporting documents via RDFa. Indeed, Diane Mueller, a member of the Board of Directors, and Dave Raggett, a W3C Fellow, a developer of the HTML specification, and one of the key people in the Semantic Web space, are currently working on such a proposal for (full disclosure: I am currently working on a book with Diane on XBRL for O’Reilly Media). Additionally, an XBRL/Semantic Activity Group has been established within the W3C in order to explore these very options, and will meet face-to-face for the first time in njunction with the annual nference in Paris in July 2009.

The Semantic Web has been moving out of the research space for a few years now, to a certain extent due to a growing understanding of the ncepts, maturity of the tools, and opportunities to apply them. XBRL is likely to be one of the more significant use cases for the technology as we move into the distributed cloud of data and all that this implies.