Written to Ralf and Kathryn:

-=-=-

Here are some thoughts (for my benefit as much as everyone else)
on how to adjust the 594 proposal so that it has a manageable
final product. My primary goal here is to get into the fall
semester without the load of an incomplete. My expectation is
that the work will continue, both as part of statistical analysis
projects for L509 and in concert with the internship with Eugene
(which, by the way, has already started if sending a big pile of
email counts).

The existing 594 proposal is relatively vague on paper:

  http://www.burningchrome.com/~cdent/uviz/L594form.doc

however the project plan

  http://www.burningchrome.com/~cdent/index.cgi?word=107

is far more detailed and grandiose. Too grandiose.

I would like to focus in on three areas, two of which are already
done:

- Migrating the data and tools from the Oracle installation on
  Ella to the MySQL installation on my own collection of
  workstations. This is done: the data and tools have been
  transfered and improved. Improvements include:
  - updated database schema to support
    - clusters
    - annotations
    - evaluations
    - more effective text searching
  - improved interface including
    - the interface to the searcing
    - efforts to allow thread traversal when reading messages
    - date based searches and retrievals

- Revising the PORT's Pragmatic Web Workship paper. This turned
  out to be _huge_ investment but also extremely valuable.
  Initially it was not included in any class, but based on the
  amount of work, I'd like it to count as credit somewhere. The
  latest version is linked from here:

    http://ella.slis.indiana.edu/~klabarre/unrev_firstpage.html

- Evaluations and annotations. These are the two areas that
  Kathryn needs. She has provided me with more detail on how
  these ought to work. I have some ideas on how to implement
  these things. What's not clear is whether the product should be
  implementing them, or only writing some design plans. That's a
  question of time.

(I've already put in a large amount of time on these things,
especially if we include associated conversations in email with
PORT people and people in the Bootstrap Alliance (the Engelbart
group). I don't want to jump ship, but I am starting to feel a
bit overburdened.)

Evaluations and anotations are divided up into three types:

- Numeric score associated with a message or group of messages
  (cluster) that indicates the relative importance of that
  message or cluster in the domain of the list. A running average
  of the scores could be kept allowing retrieval based on that
  score.[1]

- Descriptors: the attribution of pre-ordained keyword phrases to a
  message or group of messages, also to be used as retrieval
  keys.

- Annotations: writing abstracts for messages or groups of
  messages.

In terms of implementation time, the first is least complex, the
second most with the third closer to the second than the first.

That's my thoughts so far. Any comments on how to proceed?

Thanks.

On Wed, 17 Jul 2002, Debora Shaw wrote:

> Why don't you e-mail Kathryn and me your current understanding of what
> the "final product" of the L594 will be. I can attach that to the
> original form.
>
> On Wed, 17 Jul 2002 cdent@burningchrome.com wrote:
>
> > On Wed, 17 Jul 2002, Debora Shaw wrote:
> >
> > > Yes, the L596 proposal came through, and Eugene has agreed to his side
> > > of the bargain. Each of you will receive (paper) instructions on this
> > > pretty soon.
> >
> > Excellent. I'll be on the lookout for that.
> >
> > > I agree, the summer L594 description was the best guess of what would
> > > be accomplished back at the start of the term. If the proposal needs
> > > to be adapted to life as we now know it, that's fine.
> >
> > What do I need to do to make that so?

[1] Running average can be calculated as follows:

  A the most recent average
  N the number of data thus far
  x the new datum

New average = ((A * N) + x)/(N+1)