1996-02-02 - Re: noise levels

Header Data

From: Scott Brickner <sjb@universe.digex.net>
To: bryce@colorado.edu
Message Hash: b280d5555338063376b438bbb2ec09e4dd870bdb6c1b11277bea50350ada3985
Message ID: <199601312323.SAA05263@universe.digex.net>
Reply To: <199601190051.RAA28314@nagina.cs.colorado.edu>
UTC Datetime: 1996-02-02 08:40:38 UTC
Raw Date: Fri, 2 Feb 1996 16:40:38 +0800

Raw message

From: Scott Brickner <sjb@universe.digex.net>
Date: Fri, 2 Feb 1996 16:40:38 +0800
To: bryce@colorado.edu
Subject: Re: noise levels
In-Reply-To: <199601190051.RAA28314@nagina.cs.colorado.edu>
Message-ID: <199601312323.SAA05263@universe.digex.net>
MIME-Version: 1.0
Content-Type: text/plain

Bryce writes:
>Perry, I quite agree with you.  I am having a very difficult
>time wading through cpunks, and I am currently reduced to
>grepping for my name, and then picking out a topic or two by
>subject line before junking 95% of the posts.  Since you have
>such enthusiasm for solving the noise problem I suggest that we
>do the following:
<auto-kill sublist scheme elided>

I have an expansion on this.  Why not generalize the problem to create
a group rating system?  Anyone who wants to can send ratings messages
(rating each message on a scale of one to five, one meaning "what total
crap" and five meaning "what a useful piece of information") to the
ratings server.  The server maintains the ratings for each message by
sender.  Client software can retrieve the ratings added since a
given time and use this information with the ratings assigned by the
user to generate compatibility profiles indicating with which raters
the user tends to agree, and provide ratings on all messages based on
it.  The user can then have anything lower than his tolerance threshold
automatically deleted.

This is patterned after a newsgroup collaborative filtering tool I
read a paper on not too long ago.  I can't find that reference, but
<URL:http://www-sloan.mit.edu/ccs/CCSWP165.html> has an open
architecture design for a ratings server.

Ideally, one would modify MUAs to recognize an "X-Ratings-To:" header
to tell where ratings messages should be sent, and the list server
would add that to all outgoing messages.  The MUA would present
the ratings buttons when displaying messages containing "X-Ratings-To:"
headers and automatically generate and send the rating when the user
pushed a button.

The beauty of this is that it works for *any* mailing list that has
an associated ratings server.  It allows anyone with the appropriate
MUA to ignore those conspiracypunks boneheads almost transparently.

Necessary coding:
    modifications to majordomo:
	- add optional ratings server address and update frequency in
	    list configuration data
	- add "X-Ratings-To:" headers to outgoing messages in lists with
	    ratings servers
	- periodically send ratings updates to ratings server subscribers
    modifications to MUAs:
	- recognize "X-Ratings-To:" headers in incoming messages and
	    present ratings interface when displaying them
	- generate ratings messages to ratings server
	- interpret incoming ratings messages to compute user's predicted
	- maintain user preferences vector