Unlikely

Last 10 changes

peermore
peermore
peermore
aboutchris
augury
socialtext
pictures
socialtext
socialtext
aboutchris

122 words
253 defs

uvizjournal

[ Prev ] [ Next ]

Revision:
2002-08-09 02:46:56 ]
2002-08-06 03:21:50 ]
2002-07-31 17:37:12 ]
2002-07-28 03:15:13 ]
2002-07-26 14:03:51 ]
2002-07-21 17:32:20 ]
2002-07-17 14:15:53 ]
2002-07-05 23:52:07 ]
2002-07-01 22:44:44 ]
2002-06-28 17:11:51 ]

Backlinks:
unrevdb
unrevprojectplan
uvizjournal


Installed R on hot (linux) and cream (windows) to experiment
with the indexing tools provided by John. Have a question out to
him on how to take advantage of the dist() and hclust() methods
to output text files. One that contains a similarity matrix and
one that contains, in some yet to be determined format, a
representation of the clusters.

Once the unknown format is figured out, we can make the tools to
import that info and view the clusters in the database interface.

Issues:

 - how do we get the right slice out of the cluster hierarchy?
   - inter/intra sim
 - how do we track of the documents when we make the input?

This last issue is fairly important. The database represents
messages, but we may be creating documents which are made up of
multiple messages, for example dividing up the dataset by subject
or author, or date. We need to know, when we input the clusters,
what set of messages the individual documents represent.

Implies some kind of flag upon input.

Piles and piles and piles of metadata are building up.
[ Contact ] [ Old Blog ] [ New Blog ] [ Write ] [ AboutWarp ] [ Resume ] [ Search ] [ List Words ] [ Login ]