1200 Dale Avenue #100
Mountain View, CA 94040
Date: Sun Sep 16 2001 - 14:43:06 PDT
OHS DKR Project
333 Ravenswood Avenue
Menlo Park, CA 94025
Eugene's work: FAQ-index into an argument
We had a meeting with Eugene the other day, and took a look at what he had been
hammering on for a while. I have to say that I saw it in a whole new light,
which previous posts on the subject had not adequately brought to my attention.
In those posts, Eugene described making "dialog maps" of conversations. There
was a nice graphic version in PNG form, and also an HTML version.
I recall looking at the PNG. Nice, but not too useful. (Diagrams with a few
words, lines between them, but no embedded links that would let you get to the
meat of the discussion.)
I don't recall that I looked at the HTML form. THERE is where the real valuable
Those HTML pages are not HTML renditions of "dialog maps" so much as they are
FAQ-style indexes into the discussions!!
In other words, they would make a superb method for collecting and codifying
the information available in email archives, WITHOUT modifying the archives.
A few salient points woth mentioning:
- There are thousands of words in the unrev archives that may never see the
light of day, because its just to damn hard to find anything. This
mechanism could provide the answer to that problem.
- The archives are plinked (purple numbered). But Eugene has the
modified mail server to do that as messages are archived, as well as
tools to plink past archives.
- It takes a lof of effort to build the map. Basically, each day he
has to build in the information content from the day's posts. I suspect
that this will be normal -- that the addition of topic maps and
structured argumentation methodologies will require, and will in fact
produce a new profession of practicing ontologists who do this work as
their full time job. In addition:
- As users get more familiar with it, they will pre-identify
material as they create it.
- Ontologists will also find themselves re-categorizing existing
material to make it part of the archive.
- Ontologiests will be defining ontology-translations, so as to do
wholesale imports of information in other knowledge-bases
- Ontologists will monitor results produced by "ontology-
inspection agents" -- computerized programs that run around searching
for isomorphic ontologies, with an eye towards finding and mining
- QuestMap to build the dialog map. The results are then converted to an
HTML page. That mechanism probably makes it difficult for others to
massage the FAQ. But other mechanisms could be developed, and probably
will be, once the value of the result is recognized.
- One thing the system needs to do is to post links to the summarized
information. Those links would "complete the circle" -- the index has
pointers into the email archive, and the messages would have responses
that point to where the information is summarized. The most important
result of such notification pages would be the ability to click a link to
see how a discussion I have been part of was summarized -- in case I want
to make changes. (Or maybe become further enlightened.)
- The methodology is nicely applicable to structured arguments.
Especially since, as we have observed, the structure tends to become
apparent after the fact.
- One of the examples Eugene pointed to was a collection of issues
that had been discussed in a disparate collection of emails. Since
IBIS mandated that there had to be a question, he was motivated
to identify the commonality. He gathered them all under the
question: "What problems pertain to preserving link integrity",
thereby coming up with a whole new category to add to the
ontology (link integrity).
- The ontology is currently captured as a glossary. So all the
concepts are present, but relationships among them are intuitively
understood, rather than made explicit. (But, should we advance to topic
maps at some point, life may get a lot more interesting.)
- The glossary pages are maintained as T-Wiki pages, which means that
anyone can edit them, but they are access-controlled, versioned, and
there is a lot more flexibility for creating links.
In short, Eugene has come up with some good stuff that may well be the
first step in constructing a "knowledge base" out of an email archive.