I stopped by Socialtext's monthly "Wiki Wednesday" hackathon yesterday. The premise? Free pizza and beer and internet, create what you will using Socialtext's open source technology. (Disclosure note: Omidyar Network is an investor in Socialtext.) Present at the end of the day and get some sort of freebie if you've got the most popular hack. Here are the entries from yesterday:
- References. A way to add references (or footnotes) anywhere onto a page. The footnotes are parsed as wikitext, so they can structured however you like. Create a new page, and create a footnote inline as you're typing - no more difficult than marking text for italics.
- Analytics I.* There's a lot of data available from the wiki community, but not many folks have developed analysis tools. (Highly notable exception: check out what IBM can do with Wikipedia data.) In particular, how can you measure the health of a wiki community? These analytics hackers used the Socialtext REST API to suck data from public wikis, and Analysis I focused on page naming. They noted the constraint of the day limit on the hackathon, but if other wikis used similar APIs, then trends could be measured over time. (Not to mention, assessing which numbers were ultimately relevant.) So: how many linkages were engendered by page naming? Can 'link as you think' behavior could be proven using this method? In the limited analysis performed, shorter page name length was a positive factor; the popular wiki 'stoss' had a mean page name length of 21 characters, and a shorter number of tokens per page name. (* crowd favorite!)
- Analytics II. This continued wiki analytics with a review of link structures. This exercise measured the number of backlinks; the more, the better. Measuring components (which are the islands of pages that link to each other) shows that most pages don't link to each other in an unhealthy wiki. Healthier wikis have more cross-linking. The next logical step would be to take this data and compare it to Analytics I - what is the median page name length within a cluster? (* crowd favorite!)
- SocialZork by Ingy dot Net, the creator of Kwicki. This
fun-to-play-with hack was a mashup of Zork, Kwicki 2.0, and
Socialtext's REST API. Essentially, you have to earn points before you
are able to navigate off of the wiki page, make edits, etc. The gamer's
version of earning more wiki rights by building reputation.
- WhoWhatWhen. I had to miss most of this presentation, but it generally provided more sophistication around managing edit histories and versioning. (Please feel free to comment if you saw the whole demo.)
- WordCloud.* Get a sense of what people are interested in by distilling a word cloud of the 50 most popular words in a wiki. Some words that are specific to a workspace could be worth removing in order to avoid distortion - e.g., should 'socialtext' be one of the words on a Socialtext wiki? Folks in the group had active discussion on this, but no agreement to be had. (* crowd favorite!)
- Most Wanted Virtual Pages. Ended up getting a list of the most-wanted pages that don't exist, using Socialtext's incipient link feature. (Note: the creators of this hack were surprised that the results weren't coming out properly, then learned in the presentation that the incipinet link feature had been temporarily disabled! Oops.) Like a crystal ball for assessing what content site users are upset to be missing.
- Purple Transclusion. Transclusion pulls in the content from another page without actually recreating the content on the new page. This demo pulled in copy from one wiki page to another, and a click on the pulled-in content would bring you to the source content. 'Purple numbers' are used as identifiers for this content. Works just like a javascript include, plus the bonus from the click back to the source.
My favorite was the analytics hack - being able to predict or cultivate a healthy wiki community would be a terrific tool in driving adoption. I also like the concept for its potential to inform non-wiki-based online communities. Current options for online community analytics are terrible, so it would be great to see this get developed further.
Sidebar: Best t-shirt sighted at the hackathon: "GPL: Free as in herpes."