The human learning process of imitation works quietly behind everything we do. When we add computer programs to the mix we have to account for the role of the programmer.
A colleague asks, What are your thoughts on semantic wiki?
Do you use any, or achieve similar things through federated wiki? Maybe this also ties in with stuff you've learned about neo4j?
wikipedia 
A core idea from the semantic web was that words would be given precise meanings that computers could trust. Wiki leaves it to human operators to decide if the words mean anything at all.
My work with neo4j succeeds because I am careful to be sure the graph can hold contradictory information. A common use case in my day job is writing queries to understand these contradictions.
He continues, What about using semantic markup of text within a social wiki system, with something like emergent folksonomies (metadata tags), for the purpose of extending the depth of meaning for the humans (and maybe the added usefulness to the computers is just a positive externality)?
My original goal was to make a data-wiki. To that end I made sure that data could be easily represented and communicated between pages. I also made sure that anything I did could be shared openly. I worked with people who liked visualizations but each was a giant project that wouldn't be repeated soon. See Federated Wiki at One
We've had a number of years now where blogging was of more interest to users but I remain interested in casual data, where every visualization answers one question and raises two more. And, where this cycle is seen as exciting, not unrelenting drudgery.
Here is an idiom that has already emerged in Algorithmic Graphviz. See Visitor as Schema
WHERE /^When/ LINKS NODE -> HERE
This says, if you write paragraphs that begin "When" then I will draw in-links from the nodes created from links found in the paragraph.
There are actually two idioms here, one in Graphviz markup, and one in writing style of pattern language. One idiom supports the other. AND, most importantly, neither are built into wiki. Both the markup and the wiki itself is available in full to any author. Compare this to, say, Dublin Core, where you either were or weren't part of the defining committee.
Going back to my original goals, I sought to separate the collection and curation of data from the craft of building visualizations. My recent work is exactly aligned with this goal.