Semantic MediaWiki 1.2: bringing OWL to the Web

From korrekt.org

Jul 11 2008. After a lot of work it finally done: version 1.2 of Semantic MediaWiki has been released and is available for download. The new release does of course include many new features, but it also has some interesting additions regarding proper Semantic Web technologies.

In particular, SMW 1.2 now supports what we call concepts: a kind of «dynamically created categories» that are defined by storing SMW queries on special wiki pages. In this way, queries become first class objects in SMW, and the Semantic Web aficionado quickly thinks of OWL classes and gets a dreamy kind of look when thinking about the possible future of SMW. Indeed, «concept» is also what the DLista likes to call her class descriptions, and SMW now essentially allows users to edit such class descriptions in a wiki. And, of course, descriptions are exported to OWL, like in this example.

Now does this make SMW a multi-user large-scale ontology editor for OWL? Not yet. SMW 1.2 is just the first – though possibly essential – step into this direction. What is currently not possible is to organise concepts as a hierarchy, i.e. one can define classes but not assert class subsumptions. From a UI point of view this addition is of course no big step. But we also have the intention to have SMW really evaluate all the semantics that it allows people to enter. Hence there some large-scale reasoning component is still needed.

But is it a realistic goal to build such a thing, given how difficult it is to evaluate OWL ontologies? Wouldn't logical inconsistency and computational complexity thwart any attempt for getting reasoning into a wiki? I believe not. The reason is that SMW is quite careful with the expressive features it permits. With the current query language used for concept descriptions, SMW still is essentially in the DL EL++ (a well-known subset of our ELP). This part of OWL features polynomial time complexity, and it is «Horn» so that logical inconsistencies are always somewhat local (i.e.: can be pinpointed easily). This does not make the implementation easy – «polynomial» is no reason to rechoice and consider a problem solved! Extensive caching mechanisms, offline computation, and many heuristics are still required to use this in a multi-user online system.

Yet, SMW clearly is at the beginning of a path for bringing ontologies back to the web, and this certainly seems to be worth some effort.