I did not attend all that much of the Animating Archives. I did, however, manage to catch two talks by Profs. Srinivasan and Kun from California. For those of you who attended different talks, they focused on the structures of archives and how they might differ in digital media from more traditional paper-based storage archives.
Prof. Srinivasan took issue with traditional museum-based archival structures that were set in certain ontological categories. The implication was that using rigid definitions separated certain narratives from the objects in display leading to pictures that... skewed things a little.
What was important in categorizing and archiving items was what was included in their descriptions and what wasn't. In Srinivisan's case, Zuni artifacts that bore Cambridge's systemic classification system was simply inadequately capturing the full extent of Zuni culture.
Unfortunately, all universal classification systems lose something in categorization. Srinivisan's attempt at creating a Zuni archive might work - but only in a specific Zuni-based or Zuni-oriented context. Incorporating stories within archives and adding their proper cultural context (the difference between "this is what my grandmother made" vs. "for digging, stick") might help bring proper context to analysis but how far can this system go? The Cambridge system Srinivisan made an example of was probably used as a nearly-universal means of archiving and categorizing a global scenario. Srinivisan's system - a highly illuminating one - might only be applicable to very specific archives.
Only a specialist archive could get away with something so specialized. Or of a structure of archiving that is so complex. This can only work with a very specific audience that comes to the archive with a very specialized narrative in place to understand, intuitively, the necessary search parameters to access such an archive effectively. Sadly, I can't really see the addition of narratives being all that useful outside of this context.
One thing that does come to mind that might be an interesting way of looking at collecting and sifting through large amounts of data comes from my days working with intelligence services. A project was launched quite some time ago by the Pentagon's farout research wing, DARPA, that hoped to make military intelligence more efficient and eliminate some of the drudgery of analysis work that required bums-on-seats to read through countless newswire reports and field agent postings. Given the rather fluid nature of intelligence gathering, it would be very hard and counterproductive to create an efficiency division of labor that would result in easy to parse reports. What one had was a mess of intelligence redundancy - too many words on the same thing, too little on another.
And so, those boffins at DARPA and their boffin-friends from friendly overseas DARPA-esque agencies got to work creating a super Wiki. Now, I say super-wiki because what is a great big mess of intelligence reports but a collection of user-generated content that is structured with the most tenuous of categorical links? It was hoped that a brilliant new program would cut through all of that.
The new program would read the text of every report, identifying key terms like names, places, and dates, and plot them with specific phrases, generating a meta-report that identified the general premise of a subject and gave a brief summary of the entire network's data. Let's say the network has been working on "Brown University". It might pull up a set time frame's reports on the University and might link a few pertinent names and places and dates - so, for example, Aaron Wee might be linked to Wilson Hall several times over the Fall/Winter period of 2009. The report might also mention Aaron Wee was seen transferring large amounts of money to unknown sources, suspected to be Rhode Island secessionists. An intelligence analyst might get interested in this and call it a day, sending a CIA deathsquad to send Aaron Wee on an extraordinary trip. Or, seeing that Aaron Wee is not himself who she's after might choose to reorganize the search around Rhode Island Secessionists, drawing a new map where Aaron Wee is tenuously connected but might have links to a whole host of other RI rebels. Or she might pick the transfer of cash as a new search paramter and have the engine draw a new map based on that. Categories that, in of themselves, offer not much help find new meaning if they are reconfigured differently to provide new linkages. The same principle sort of works for the real Wikipedia too - who hasn't got caught lost on the internet, randomly hyperlinking from wiki page to wiki page in a mad mad tour of the world. Categories, despite not carrying narratives in of themselves, still ride on the internal narrative of the searcher. It is in how one views the world that categories are created - archival categories sometimes just make those easier to spot.
Having discrete categories does help in searches. Having them broad does stifle perceptions. But if categories are linked well and then re-fashioned to identify new patterns, they aren't so bad after all.