the reflexions insider

Unless you follow the often arcane world of web standards and the organization that defines them, you probably haven’t heard much about two emerging standards that by some credible accounts will do more to change the landscape of the internet than anything since the web itself: RDF and SPARQL.

Part of what makes these technologies so compelling is their elegance: they are incredibly simple and yet have the potential to yield extremely powerful and useful results. (NB: while this entry focuses on RDF and SPARQL, these are only two pieces of a larger heirarchy of standards and frameworks designed to help realize the semantic web as a widespread and practical technology.)

RDF, or Resource Description Framework, is a collection of methods for structuring information and representing knowledge. The goal is to enable applications to automatically catalogue and exchange information based on common sets of meaningful terms, or ontologies. In its simplest form, RDF statements are triples of the form (a,b,c) representing subject-predicate-object statements. It’s a subtle but powerful tool that can enable the efficient sharing of facts — e.g., (”Kinshasa”,”isCapitalOf”,”Democratic Republic of Congo”) might be a representation of a fact about an African nation’s capital (note that the same triple can be used to derive information about both the subject and object). (In fact rather than a string, each element of the triple is actually something called a URI, which is a pointer to a resource much like a web address.)

Incredibly large and queryable databases of these triples, or so-called “triple-stores” are starting to be compiled by public and private institutions and the resulting possibilities are humbling.

Some early examples of applications that utilize these standards give a taste of what’s possible. Semantic MediaWiki, an extended version of the popular wiki platform, enables quick and powerful searching of entities and their relationships. And a combination of resources like social networks, GIS databases, and the widespread availability of development frameworks have begun to enable developers to build so-called “mashups” that are only starting to gain recognition.

RDF and these vast triple-stores will derive their power from two related sources: shared ontologies (e.g., vocabularies of terms with agreed-upon referents), and a SQL-like language that can be used to perform queries across these data sets. A strong contender for such a language, SPARQL, is still being refined by the W3C but has already been widely adopted.

But there are some roadblocks preventing RDF and SPARQL from changing the world in ways envisioned by their most ardent advocates in the immediate future. One of the most interesting of these is the shared ontology problem. Indeed, building a shared database of knowledge when the terms in your vocabulary are in dispute is a problem without an obvious answer (consider whether the term “Palestine” refers to any clearly-defined region or political entity, for example).

One discovers that the holy grail of knowledge representation is also subject to what is both the fundamental enabler and challenge of human discourse: widespread agreement.

You must be logged in to post a comment.