Semantic MediaWiki

I’ve been aware of the Semantic MediaWiki project for some time. Today I finally had a closer look at their demonstrator, and I have to say I like it.

Here is an example page about San Diego. The interesting part are two boxes at the bottom of the page: “Relations to other articles” and “Attributes of San Diego”. They show formally captured, machine-processable facts from the article.

The wiki markup of the page reveals that both kinds of statements, relations and attributes, are captured by bits of inline syntax.

Relations:

'''San Diego''' is a [[is a::city]] located in the southwestern
corner of [[is located in::California]], the extreme southwestern
corner of the [[located in::United States]].

Attributes:

As of [[2005]], the California Department of Finance estimated
the city to have [[population:=1,305,736]] residents.

Each page has a “View as RDF” link. The linked RDF/XML (example) contains all statements made on the page, as well as RDFS descriptions of all types and attributes used in these statements, which is a good idea. All terms are placed into the same namespace. There’s no way to import or relate terms to foreign namespaces.

There’s a semantic search/query feature, accessible through the magnifier icons. I don’t care too much about it. I can just crawl the wiki and load the RDF/XML representations of all pages into my own triple store for querying or building alternative user interfaces. This is a greater benefit than the improved search, I think.

All of this integrates very well with MediaWiki’s slick and familiar user interface.

I would love to see this in production at Wikipedia.

Update: A bit of semantic wiki history and a look at the bigger picture [in my followup]