Re: Gnus and GroupLens

[bmiller@cs.umn.edu (Brad Miller)]

>>>>> "David" == David Allen <davida@lsl.co.uk> writes:

>>>>> "Brad" == Brad Miller <bmiller@cs.umn.edu> writes:
>>>>> "David" == David Allen <davida@lsl.co.uk> writes:

David> Well, it says "Sending Ratings...Done" when I quit the group.  I don't
David> know that I can check whether they are actually getting there.

The trace buffer is called trace of BBB session to grouplens.cs....
you can look at this buffer and see if you got an
OK
receved messages

at the bottom, after your list of message-ids and ratings.

Brad> The BBB has a three hour inactivity timer on your login
Brad> token.  If you haven't sent any ratings or received any
Brad> predictions for more than 3 hours, you should see a message
Brad> telling you to log-in.  Just M-x bbb-login will do the
Brad> trick.

David> OK.  The symptoms I were seeing involved a message that mentioned
David> timeout that was very quickly overwritten so I wasn't sure what had
David> happened.  I'll check this out next time it times out.

David> Interestingly, the first day I started using GroupLens, most of the
David> articles in the GroupLens groups had predictions.  Today, none of the
David> new articles seem to have predictions (I can still see the predictions
David> for old articles).

David> Presumably, with GroupLens you should be able to read only articles
David> with high predictions (plus any you might select for other reasons).
David> You will then only be rating a relatively small number of articles.
David> Do you expect that with a large user community, each user will only
David> need to rate a very small percentage of articles but there will still
David> be enough ratings for the system to work well?

That is the theory.  For really high volume groups, I expect that
GroupLens will need at least a hundred people reading and rating, to
keep a good mix of rated articles.  A thousand would be even better.

Some days are definitely going to be better than others.  For example in
comp.lang.java we had a couple days in a row where there were over 500
ratings entered.  I just checked the stats now, and there were only
16 ratings in the last 24 hours.  You can check these stats off the
GroupLens homepage http://www.cs.umn.edu/Research/GroupLens/stats/daily.html

Getting GroupLens past the critical mass point seems to be our most
challenging problem right now.   There are several issues involved here:
   - Publicity
   - The number of supported newsreaders (including the total lack of a
     Windows newsreader)
   - The ability of the average user to install a special version
     of a newsreader just for GroupLens 
   - Getting users to change their behavior, and remember to rate the
     articles as they are read.

I'm sure there are other factors at work too.  I'm open to any ideas on
how to get more people using GroupLens.   

\Brad
----------------------------------------------------------------------------
Brad Miller                    | e-mail: bmiller@cs.umn.edu
University of Minnesota        | phone:  (612) 626-8396     
Department of Computer Science | www:    http://www.cs.umn.edu/~bmiller
EE/CS 5-244                    | 
----------------------------------------------------------------------------