http:^^www.cs.umn.edu^research^grouplens^glfaq.html

This data set contains WWW-pages collected from computer science departments of various universities

Date: Tue, 14 Jan 1997 22:14:23 GMT
Server: Apache/1.1b3
Content-type: text/html
Content-length: 11139
Last-modified: Sat, 27 Jul 1996 19:47:00 GMT

<center>
<H1>Frequently Asked Questions about GroupLens</H1>

<!WA0><IMG ALIGN=center SRC="http://www.cs.umn.edu/Research/GroupLens/at_work.gif">

</center>
This FAQ is maintained by Brad Miller  <!WA1><a
href="mailto:[e-mail]."> [e-mail]</a>.  GroupLens is
my PhD thesis project.  The opinions expressed herein belong to me.<br>

Last Update:  April 11, 1996

<H2>Table of Contents</H2>
<OL>
<LI> <!WA2><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#WhatisCF">What is Collaborative Filtering?</A>
<LI> <!WA3><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#WhatisGL">What is GroupLens?</A>
<LI> <!WA4><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#WhyGL">Why should I use GroupLens?</A>
<LI> <!WA5><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#Whynotall">Why Doesn't GroupLens support all newsgroups?</A>
<LI> <!WA6><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#WhatNG">What Newsgroups are supported?</A>
<LI> <!WA7><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#howtoaddNG">How can I convince you to add my
favorite newgroup?</A>
<LI> <!WA8><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#WhatNR">What Newsreaders are supported?</A>
<LI> <!WA9><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#newsreaderno">How can I convince you to add
grouplens support to my favorite newsreader?</A>
<LI> <!WA10><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#longterm">What are the long term plans and
goals of GroupLens?</A>
<hr>
<LI> <!WA11><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#pseudonym">What is a pseudonym, and why do I need one?</A>
<LI> <!WA12><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#whatpred">What is a prediction, How is it calculated?</A>
<hr>
<LI> <!WA13><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#nopred">Why don't I get any predictions?</A>

<LI> <!WA14><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#e502">What is 502 You have no permission to talk?</A>

<LI> <!WA15><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#howratereader">How Do I enter ratings in gnus/tin/xrn?</A>

<LI> <!WA16><A HREF="http://www.cs.umn.edu/Research/GroupLens/glfaq.html#howratedecide">How Do I decide what to rate
an article?</A>



</OL>

<H2>The Answers</H2>
<A NAME=WhatisCF><b>What is Collaborative Filtering?</b></A><br>
 Collaborative filtering is a technology that is designed to deal with
 the problem of information overload.  
 Collaborative filtering is made up of a number of technologies for
 gathering ratings from an on-line community and transforming them into
 recommendations.  Collaborative filtering systems exist for helping you
 select audio CDs, video tapes, and even restaurants.

<p>
<A NAME=WhatisGL><b>What is GroupLens?</b></A><br>

GroupLens is a collaborative filtering system that helps you work
together with other people to find the quality news articles out of the
huge volume of news articles generated every day.
<p>
To accomplish this the GroupLens system combines your opinions about
articles you have already read with the opinions of others who have
done likewise and gives you a personalized prediction for each unread
news article.  Think of GroupLens as a matchmaker.  GroupLens watches
how you rate articles, and finds other people that rate articles the
same way. Once it has found for you some people you agree with it tells
you, in the form of a prediction, what they thought of the article.
You can use this prediction to help you decide whether or not you want
to read the article. 
<p>
<A NAME=WhyGL><b>Why should I use GroupLens?</b></A><br>
If you no longer find reading Usenet news to be an enjoyable part of
your day, you should try GroupLens.  I beleive that colaborative
filtering, and GroupLens can make a fundamental change in the way we all
use Usenet.  It seems that on almost every active newsgroup today, there
are periodic discussions, and calls for either moderation, or
reorganization.  Why does this happen?  Too much junk, too much volume,
and no sense of community.
<p>
I believe that collaborative filtering is a better solution
than newsgroup reorganization or moderation.  The arguments against
reorganization are well known:  Reorganization leads to lots of
crossposts when people aren't sure where an article belongs.
Reoganization fragments the community and forces the user to go looking
for information in many different locations.  Moderation, whew!  What a
job to try and moderate 400 -- 500 new messages each day.  Plus, what if
I don't like the moderator?  What if he and I don't have the same
interests or the moderator has his own agenda for the newgroup?
Collaborative filtering can give us the best of both worlds by
democratizing the moderation process.  With GroupLens, we can all be
moderators for each other.
<p>
Right now, GroupLens is still in the "early adopter" stage.  That is,
only a few hundred hardy pioneers are using the system.  But, one of the 
interesting characteristics of collaborative filtering systems is that
the more people that use a system, the better the system works.  This
goes double for Grouplens!  All of the groups that I have received
requests to add GroupLens support for are high volume high noise
groups.  If everyone only wants to read and rate a few articles each
day, then it takes lots of people to cover and provide ratings for all
the articles.

<p>
<A NAME=Whynotall><b>Why don't you support all newsgroups?</b></A><br>
We cannot support all newsgroups right now, because the volume of data
would be too huge, and the network traffic would be far to great for any
one site to handle.  Usenet is a huge distributed system.  Trying to
centralize ratings of every article in the system would be crazy.
<p>
One of the goals in the early stages of GroupLens, is to understand just
what the critical mass of users is to make GroupLens work very
effectively.  Once we've defined the parameters, we can distribute the
GroupLens and then we'll be able to handle all the newsgroups.

<p>
<A NAME=WhatNG><b>What Newsgroups are supported?</b></A><br>
<ol>
<li> All of comp.os.linux.*
<li> rec.humor
<li> rec.food.recipes
<li> rec.arts.movies.current-films
<li> comp.lang.c++
<li> comp.lang.java
<li> comp.groupware
<li> comp.human-factors
<li> mn.general
</ol>

<p>
<A NAME=howtoaddNG><b>How can I convince you to support my favorite newsgroup?</b></A><br>
You must bring me some users!  Thats it.  If you really want to have
GroupLens support your favorite newsgroup, just find some other people
on that newsgroup who are willing to read and rate the articles.  If you
can give me a list of users who will promise to use grouplens, I'll add
support for it.  The higher the volume group, the more users I'll expect
you to find.  But how, how do I find these users you ask?  I don't
know.  I've been a GroupLens  evangelist for the past eight weeks and I
don't have any magic answers.  Some people have responded to my posts to
Usenet describing the project with great enthusiasm.  Luckily hardly
anyone has told me this is a bad idea.  Netiquette would seem to dictate
that one post to your favorite group saying something like "Join me in
GroupLens" is OK, especially if you direct followups to news.groups.
<p>
<A NAME=WhatNR><b>What Newsreaders support GroupLens?</b></A><br>
<ol>
<li> gnus-5.x, the mother of all newsreaders!
<li> tin
<li> xrn
</ol>

<p>
<A NAME=newsreaderno><b>How can I convince you to add GroupLens support to
my favorite newsreader?</b></A><br>
I know, there are hundreds of newsreaders out there that do not have
GroupLens support.  It would be especially great to have a Windows,
and/or Macintosh newsreader on the list above.  We're working on some,
and could really use some programming help.  Its not hard to add
grouplens support to a newsreader.  Especially if you're already
familiar with the internals.

<p>
<A NAME=longterm><b>What are the long term plans and goals of the
GroupLens project?</b></A><br>
Long term, GroupLens should support every newsgroup.  This is going to
mean that we have a network of ORBs all working together in a distributed
system.  If you or someone you know might be interested in running a ORB
at their site, please contact me.

<p>
<A NAME=pseudonym><b>What is a pseudonym, and why do I need one?</b></A><br>

A pseudonym is your identifier to the Open Ratings Bureau.  GroupLens lets
you choose your own pseudonym for your ratings and predictions.  Noone
-- not even the GroupLens server -- knows who "SpacemanSpiff" *really*
is.
<p>
In the future, we'll be improving the security model of the ORB even
more so that you can use a password along with your pseudonym so that
nobody else could use it.

<P>
<A NAME=whatpred><b>What is a prediction, how is it calculated?</b></A><br>
A prediction is grouplens best estimate of how much you will like a
particular article.  To calculate a prediction you need two things.
First you need some measure of how similar you are to the other people
who have read and rated articles.  Second you need at least one (and
preferably several) ratings for the article you want a prediction for.
GroupLens combines these two pieces of information for all the available
ratings for this article to come up with a prediction.

<P>
<A NAME=nopred><b>Why don't I get any predictions?</b></A> <br>
The most common reason for not getting predictions is that you haven't
entered any ratings.  In order to calcluate a prediction you need to
have rated some articles that other people have also rated.  Another
reason why some articles are not rated is that you are the first
GroupLens user to rate the article.  The nice thing about GroupLens is
that the more people we get to participate, the less frequently that
will happen to you.

 <P>
<A NAME=e502><b>What is 502 You have no permission to talk?</b></A> <br>
This is an nntpserver error message.  The most common reason you will
see this message is that the binary version of the grouplensified tin,
has newsstand.tc.umn.edu compiled in as the default news server.  This
is the newsserver for the University of Minnesota, and will not serve
news to anyone outside the umn.edu domain.  The way to fix this is to
set your NNTPSERVER environment variable to point to your nntpserver.
Or set your nntpserver through the tin configuration menus.

 <P>
<A NAME=howratereader><b>How do I enter ratings in the xxx newsreader</b></A> <br>
The answer to this question can be found in the HOWTO use GroupLens
with xxx document.  
<ul>
<li> <!WA17><a href="http://www.cs.umn.edu/Research/GroupLens/software/gnus.html">
http://www.cs.umn.edu/Research/GroupLens/software/gnus.html</a>
<li> <!WA18><a href="http://www.cs.umn.edu/Research/GroupLens/software/tin.html">
http://www.cs.umn.edu/Research/GroupLens/software/tin.html</a>
<li> <!WA19><a href="http://www.cs.umn.edu/Research/GroupLens/software/xrn.html">
http://www.cs.umn.edu/Research/GroupLens/software/xrn.html</a>
</ul>
In general entering a rating is as simple as
typing a number from one to five.  Sometimes the number must be
followed by another key, and sometimes not.  The problem with most
newsreaders is finding a keybind that isn't already used for something
else! 

 <P>
<A NAME=howratedecide><b>How do I decide what to rate an
article?</b></A><br> 
Each user will have their own criteria for how they rate different
articles.  We suggest something like the following:
<OL>
<LI> This article is a waste of net.bandwidth, I definitely don't want
more like this.
<LI> This article is bad
<LI> This article is OK
<LI> This article is good
<LI> This article is great, I'd like to see more like this one!
</OL>
If you would be willing to participate in a survey at some time about
how you rate articles in GroupLens, please send mail to:
<!WA20><A HREF="mailto:[e-mail]">grouplens-admin</A>