You are hereSlide / Published Subjects: Small Pieces, Meaningfully Joined

Published Subjects: Small Pieces, Meaningfully Joined


By murray - Posted on 04 April 2008

Modern web sites have demonstrated the utility of keyword based tags as very lightweight indicators of a particular subject to human viewers, e.g. to indicate what a video posted on youtube.com is about, seen from the perspecive of the poster. In addition, algoritms can use identical tags within the same address space as string based identifiers to manage sets of objects or data presumably about the same indicated subject, e.g. “person”.

Many sites have also made these sets globally adressable over HTTP, as in http://flickr.com/photos/tags/person/, which, combined with common formats like RSS and Atom, make it technically and syntactically straightforward to merrily aggregate sets of information items across multiple sites. …Simply assume that the identifiers for the different data sets identify the same subject – and presto…! Similar set based operations, even updates, can be performed across other distributed information systems, given a shared address space, protocol, and format(s).

The fact that interoperability at this level is technically feasible, however, doesn’t guarantee that any given set operation is in any way meaningful. In some contexts, such operations can be illegal – even dangerous. E.g. the well-known FOAF information model says that ”Something is a foaf:Person if it is a person. We don’t nitpic about whether they’re alive, dead, real or imaginary”. It is obvious that imaginary persons could be risky to involve in e.g. medical or legal contexts. From the discussion at http://en.wikipedia.org/wiki/Person, it should be similarly obvious that different definitions of “person” are needed in different contexts.

While diversity is good, it becomes a challenge when e.g. a quick search for “localname:person” at the “semantic web search engine” Swoogle at the point of this writing returns 1873 different URI-based identifiers, and provides little or no information on how these identifiers are applied in various contexts, how to stay updated on changes, and how to understand various conceptual model related constraints and assumptions. How is a poor human to know which one to reuse for meaningful (semantic) interoperability?

Based on the well-established “Levels of Conceptual Interoperability Model (LCIM)” from Virginia Modeling Analysis & Simulation Center, this presentation will go some way towards indicating why identifiers in distributed environments should not only resolve to provide natural language semantic disambiguation where necessary, but also provide services for further discovery of application context specific information to ease reuse. Finally, Enterprise Architects will be interested to know which business contexts appear to benefit most from the dynamic, model driven approach to interoperability suggested in this presentation. Recent findings from the European Commision/IST funded “Informal Study Group on Value Proposition for Enterprise Interoperability” will be presented briefly to give some indication on this.