Word of e-mouth
(we're trying to find a name for the project)
For a person, like myself, who lives on
the Internet and profits by the connections I can make within the information
that flows at me, the time between conception of something and finding
it becomes critical. Search engines are helpful but too much time
is spent weeding and figuring out key words. This project attempts
to create a tool that will evolve to solve that problem.
My solution does not to involve building
another search engine web site. Rather it involves building a personal
information organizer which has access to all the information that flows
to me (emails, usenet, web pages, chat dialogues, word documents etc.).
This personal organizer (PO) also has access via a peer-to-peer
(chat) facility to all my contacts on the internet and consequently to
the information in their POs. Since it is relatively
easy for each of us to characterize our interest it will be relatively
easy for us to create SIG groups and publish the lists of members to everyone
with a PO. Consequently I should be able to have a chat dialogue
with a SIG group at large rather than a specific person. I should
be able to retrieve URLs from other's PO whether or not they are present
at their computers. What might emerge, then, would be a kind of distributed
I figured for this to work, the PO would
need to index my emails, usenet postings, and the web pages I bookmark;
filing them such that they could be retrieved by me (or one of my
contacts) later by their context alone. If I bookmark a page about
'latent semantic indexing' and then
6 months later type in something like 'information retrieval techniques'
ideally I would want that hyper link (and others) to suddenly appear for
me to click on. The same would apply for all my contacts - though
perhaps somewhat delayed. I realize that this is a tall order,
and i don't expect results like that immediately, but I think that a peer-to-peer
chat facility connected to a personal organizer connected to an advanced
method of information indexing and retrieval will provide the best platform
in which such a facility can emerge.
From the site AUTOSPEC
THEMATICS: KEYS TO CONTEXT AND CONCEPT FILTERING we get the idea
that the vocabulary of a special interest group (SIG) is the best background
for relevant retrieval of information. In my scheme the SIG vocabularies
and references to texts based on those vocabularies are passed from peer
to peer. As such, that information becomes the shared social memory
of the group.
The nature of the project
The first stage for me is to get the big pieces
into their correct relationships so the project will have a chance to survive.
By big pieces I mean social, political, and economic things.
Choosing to rely upon the social connections of chat groups and providing
people with a tool to remember their own subjective information in the
process of getting information from others, should set the stage for something
interesting to happen. I see an open source project where the basic
personal tool is given away. It should evolve into the ideal platform for
programmers to test out things like NLP intentional agents. For that
we need to open up a sand box that starts out doing some immediately useful
information organizing tasks and provides vocabulary, indexing, and internet
communication tools. I'm thinking that if those ingredients are present
this thing should take on a life of it's own. So the project will
be grass roots, bottom up, and fueled by people's natural interests and
demands for knowledge, rather than top down and fueled by capital and control.
But that doesn't mean that we can't make money in the process. I
see forming a corporation early on that is destined to be publicly traded.
When people contribute in a practical way to the project , what they get
out is stock, knowledge and contacts. Of course getting things
to actually happen that way is going to be the fun part. If
that wasn'twhat you had in mind, I guess it's time to bail out.
The User Interface
A separate application that is designed
to always be active on the computer. It should have access to all
the kinds of information objects that flow in and out of the computer from
intentional actions of the user. Patrik listed these, let me rephrase
them as follows: (note that these are all primary info sources intentionally
accessed by the user)
It should also have access to the search engines
on the Web just like Zurf does, as well as peer-to-peer connections with
all other similar systems on its contact list. It can make
searches and and communicate with other users any time the computer is
idle. It has a standing order to enhance it's connections to the
contexts of the objects that have been intentionally given it by its user.
email (read or written)
usenet articles (read or writteen)
chat sessions (if saved)
web pages visited
web pages bookmarked [our system replaces
and/or enhances IE favorites and Netscape bookmarks]
other word files read or written (.doc etc.)
3 major areas of the GUI screen:
Note that these three areas look similar to
the major areas of windows explorer or Zurf or a news reader. The
major difference from these other GUIs is that context is defined by a
SIG vocabulary (< 50 words) and not by a heretical tree. Objects
for display in the RA are based upon the mouse clicks inside the
SIGVL and not by navigating a heretical tree of directories or folders.
Hmm ... some more though needs to happen here re moving objects from context
to context ... more dialogue please ....
Input box (IB)- this is where you type
in your queries, comments, or questions. Input is always assumed
to be in the context of the last object that was selected unless the user
explicitly opts to find a new context.
SIG vocabulary list (SIGV)-
this area is also always assumed to be in the context of the last object
selected. Each word would have a check box next to it showing the
state of that concept relative to the object of current interest.
When the context object changes the system shows (or guesses) the SIG vocab
for that object and the state of each word relative to that object.
The states would be something like [yes, no, don't care]. The user
can toggle the state of each word by a click of the mouse. Every
mouse click is a vote on that word in the global SIG, and a definition
of the that word relative to the current context object in the subjective
view of the user. Each mouse click can also be taken as a reward
to every agent that contributed to guessing that state correctly and a
punishment to all agents that guessed incorrectly.
References area (RA)- This should look
like a jump list (book mark list) of the current context. It is the
POs output area to the user. Every message, question, or pointer
irelivant to the user's specified context is displayed here. Each
entry is clickable and opens in the appropriate window: browser, email,
word processor etc.
Revised 4/29/2000 by Seth
Mailing list eGroups