[1008] in magellan
KB Discovery Project Status Report -- June/July 2003
daemon@ATHENA.MIT.EDU (smyser)
Tue Aug 5 12:30:24 2003
Reply-To: <smyser@MIT.EDU>
From: "smyser" <smyser@MIT.EDU>
To: <magellan@mit.edu>
Cc: "'Knowledge Base Discovery Project'" <kb-project@mit.edu>
Date: Tue, 5 Aug 2003 12:30:18 -0400
Message-ID: <005601c35b6e$dbc16280$20029812@soprclar>
MIME-Version: 1.0
Content-Type: text/plain;
charset="us-ascii"
Content-Transfer-Encoding: 8bit
June/July 2003 Status Report
Project name: Knowledge Base Discovery Project
Project leader: Rob Smyser
Web URL: http://web.mit.edu/is/discovery/kb/
Accomplishments past period:
- integrated survey responses from the ct-lead and it-partners community.
The upshot is that people want an Apple-like interface on top of an AFS-like
file store with revision control.
- developed a comprehensive features list in Excel; useful in tracking
vendor features during the demos.
- Had in-depth demos/reviews of these vendors:
A.) KnowledgeBase.Net
-- an outstanding system in terms of design, but hopelessly stuck on
Microsoft-only technologies. A candidate for reverse-engineering but not
for acquisition. It really is a pity.
B.) ServiceWare -- the web-selfservice module provides to the do-it-yourself
customer the same problem resolution experience they'd get in a live call
with a helpdesk agent, if that agent was using Serviceware to help solve the
problem. We found it a little too tightly focused on that one style of
problem-solving. Move on.
C.) Kanisa -- the system that is the Apple interface. MySupport and Guided
Search are outstanding and compelling. The system is NOT a content
management system, though you can author in it; instead it machine-reads all
your documents and absorbs them into a conceptual structure based on
linguistic understanding of the content. The key win here is that it allows
for a highly-distributed federated environment (such as MITs) by installing
a layer of "understanding and guidance" on top of the existing silos of
content vetting and preparation. This feature makes it truly
enterprise-quality as you don't have to convert your content or your
processes in order to participate in the system.
- drilled only dry holes looking for open source knowledge bases,or systems
that bill themselves as such.
- searched around for and installed as demos several Open Source content
management systems. Looked at WebGUI, PostNuke, OpenCMS. Nothing so far
has been quite right. Keep looking (for a while).
- established a dialog with Indiana about their KB. They set up an instance
of their system for us for $$. Working on a mutual-aid pact sort of
negotiation. We'll have to see.
- established contact with Princeton. Turns out their KB and
problem-tracking system is freely available and is compatible with our
enterprise standards (to wit, Perl and Oracle). Gonna pursue this further.
- looking at RTFM, the Faq manager add-on to the RT system that the Tooltime
team is looking at as a possible swap-out-the-engine for Casetracker. A
young, 0.9 version, but extensible. It would be nice to use the exact same
accounts and security system as CT. But that does limit it more tightly to
the "problem-solving" domain of KB, which isn't the whole enchilada.
Reserve judgement.
Goals for the coming period:
- reach some conclusions on the strategic path to take with the KB
recommendations.
- somehow get a demo of RedHat CCM and Portal, the most promising of the
open source CMS systems.
- continue negotiations with Indiana.
- set up and test Princeton's KB.
- pursue the idea of a consortium arrangement with Stanford on Kanisa.
- begin to fill in the report.
- document the typical people architecture of a university KB setting,
talking with Duke, Indiana, and Princeton.
Next community milestone:
None planned
Issues:
It's possible that we will find no candidate in the open source domain.
Key learnings:
- princeton shows that interesting things might show up at the last minute,
and we to be prepared to consider them quickly.
Team dynamics:
Additional comments:
