Back on 25 July 1996 a friend (CA) and I were having an online conversation about how easy it is to propose, and hard it is to actually implement, a "knowledgebase" — a then-popular buzzword for a shared repository of community wisdom. My comrade had pointed out how much expert work has to be done in order to build and maintain such a repository. In violent agreement, I noted:

"Most of the promising db (sorry, I can't say "knowledgebase" today!) projects I've seen have run aground on several shoals:

  • Procrustian structures for the db contents — e.g., one or a few hierarchies, maybe based on people, places, projects, and various topical keywords — so when new, unanticipated questions come along which cut across the categories, the database reverts to being (at best) a flat file. The dearly departed R— S-----—, D-— W—, and others had some ideas on how to attack this problem; seems to me to be nontrivial (to put it mildly!).
  • Lack of time — insufficient investment by the substantive db builders and maintainers, who are going to have to spend at least a few (10? 100? 1000?) seconds per input item, in order to flag terms and add value (links, identifiers, descriptors, key words, commentary, etc.) beyond the raw text.
  • Noise — the bane of the current WWW, the lack of quality control and accountability and stability of db contents ... difficulty of separating wheat from chaff, of tracking information back to its ultimate source, etc.
  • Non-Scalability & Hubris — plans to revolutionize life, based on clever/promising AI approaches that worked great for narrow 'toy' problem areas, but don't scale up to big, complex, heterogeneous, redundant, messy data streams such as those we have to wrestle with in reality."

Five years later, are there ways around these problems? I hope so — the computer tools have arguably improved — but without considerable human expert help, it's still a tough nut. For some aspects of the problem space I'm beginning to think that Wiki concepts can perhaps help ... see The Wiki Way and/or [1] for starting points. More on that another time!

(slightly edited from the original; see LearningInvestment, ^zhurnal 11 February 2000, and ZhurnalWiki Preview)

Tuesday, April 17, 2001 at 05:58:57 (EDT) = 2001-04-17

TopicProgramming - TopicOrganizations

(correlates: TheWikiWay, GoodNotation, IdeaChampions, ...)