web page hit counter

Monday, May 01, 2006

First Monday, GroupWare and Latent Semantic Analysis

For hard core geeks, it's fun to play the "follow the definition" game with academic papers, looking up references (hopefully canonical, often not) and following them backwards (or across) until you either arrive at the tenuous source of an idea, or are totally lost, your metaphorical canoe all tangled up in the mangroves.

I burned some cycles over the past couple of days following references from an article in the most recent issue of First Monday[1]:

Beyond binary choices: Understanding and exploiting trade-offs to enhance creativity
Gerhard Fischer, Center for LifeLong Learning & Design (L3D) , University of Colorado

First Monday is right up there with Citeseer[2] on my list of fun stuff for the academically oriented practitioner. The articles that aren't actually good are often unintentionally funny[3] so I get some enjoyment regardless.

The "Binary Choices" article discusses the tradeoffs and advantages in creating end-user extensible systems, with a focus on collaboration. It hooked me with this quoted quote from the list of research objectives:
... empowering users rather than replacing or deskilling them by emphasizing knowledge based support environments from an intelligence augmentation perspective (Bobrow, 1991; Fischer and Nakakoji, 1992)
Very Engelbart-ish[4], and a refreshing change from the "Thanks for writing this great collaboration tool! Now I can outsource your job!" attitude that seems to have replaced much of the original vision behind the modern workstation.

The article was fun, but I was tantalized by hints of a computer scientist hiding beneath the humanities-slanted verbage and decided to follow the trail back a level to one of the author's previous First Monday articles.

Beyond "Couch Potatoes" : From Consumers to Designers and Active Contributors
Gerhard Fischer, Center for LifeLong Learning & Design (L3D) , University of Colorado

I'm a skeptic when it comes to turning couch potatoes into active participants. Sometimes I like being a couch potato and I resent attempts to convince me that sitting down for the occasional MythBusters marathon is somehow wrong. But there was an undercurrent of practicality in the the article, including this very nifty quote:
Environments inviting informed participation are created with many good intentions at design time but the participation and contributions that actually occur often do not live up to the expectations, a finding which can be characterized by "build it and and they will not come" (Smith and Farquhar, 2000). Analyses of these failures of adoption and sustained use of systems can be seen in many domains, including groupware applications, organizational memories, and knowledge management systems.
Which made me think about nazi-ware and failure. "Nazi-ware" was how users fondly referred to early-90's group collaboration applications. Much as the Semantic Web is rehashing the failures of the mid-80's Knowlede Representation fad, the Social Networking craze seems to want to repeat the horror-that-was-GroupWare (but this time they're asking for volunteers). Fischer offers a measuring stick for how bad things are:

Utility = Value/Effort

Putting numbers to the equation would make for an interesting exercise, but the article doesn't attempt it. I surfed on over to the author's home page (not strictly keeping with the rules of the game) and found:

Articulating the Task at Hand and Making Information Relevant to It
Gerhard Fischer, Center for LifeLong Learning & Design (L3D) , University of Colorado Human-Computer Interaction Journal, Special Issue on Context-Aware Computing, 2001 Volume 16, pp. 243-256 http://l3d.cs.colorado.edu/~gerhard/papers/hci2001.pdf

Which was worth reading, but is mainly relevant because it mentioned LSA (Latent Semantic Analysis), which sent me off on a tangent to figure out why that buzzphrase was popping up in my life so often. I'd been vaguely interested in it since Bar Camp Dallas[5] so I slid across to:

A Solution to Plato's Problem: The Latent Semantic Analysis Theory of Acquisition, Induction and Representation of Knowledge
Landauer, T. K. & Dumais, S. T. (1997)

I don't care if LSA is linguistically realistic, or even if it works, I just think it's cool that somebody managed to mix singular value decomposition and semantic search. Linear algebra, what can't it do?

So I think I may need to sprinkle a little LSA on Event Mirror and see what happens. Two things are becoming obvious: (a) Time as an axis isn't quite enough and (b) Starting out with a first-pass organization method, then allowing refinement, is better than requiring the user to do everything from scratch.

[1] http://firstmonday.org/
[2] http://citeseer.ist.psu.edu
[3] Ninety percent of everything is crap, so don't take it too seriously. One of the articles in the current First Mondy issue explains negative feedback on a groupware application by blaming "the rampant ideology of capitalism and competition".
[4] http://sloan.stanford.edu/mousesite/1968Demo.html is the best demo, of anything, ever. Stop reading this and go watch it.
[5] Latent Semantic Analysis/Search, David Hagar, Pure Discovery.
http://barcamp.org/Dallas2006Schedule, 3:30-4:00 in the Mash Pit room. Unfortunately not recorded.

tags: , , , , , , , , ,

You should follow me on twitter here.


Blogger daniel said...

brilliant. will have to explore when less sleep and time deprived.

8:34 AM  
Anonymous Jay Fichialos said...

I'm interested in why you think time (or should I say an individuals perception of time) is not enough to serve as an axis? Sure it’s not an absolute in the sense that it’s unchanging, but I still think an individuals perception of time and events can serve as a valuable axis from which to build a sense on context and shared understanding… Fantastic write-up, enjoying the trail.

6:13 AM  
Blogger cks said...

Time's good, but even just the IRC chat logs from #cocoadevhouse are big enough at this point that it's hard to follow without some extra level of organization. And the Email stream (I've got microformats-dev hooked up now) is pretty much totally useless without some extra organization. Search is good, but I was thinking some sort of auto-tagging (based on LSA) might be fun. Well, some day, lots of other stuff to do before that.

7:38 PM  

Post a Comment

<< Home