NAME
clusterlabeling.pl Label discovered clusters based on their content
SYNOPSIS
Assigns labels to each cluster with the significant word pairs found in the cluster contexts. Also separates the clusters in different files. This is particularly useful for the web-interface.
Two types of labels are assigned to each cluster : Descriptive and Discriminating. Descriptive labels are the top n significant word pairs. Discriminating labels are the word-pairs unique to the cluster out of the top n significant word-pairs for the cluster.
USAGE
clusterlabeling.pl [OPTIONS] INPUTFILE
Required Arguments:
INPUTFILE
File created by Toolkit/evaluate/format_clusters.pl with --context option.
Optional Arguments:
--token TOKEN
A file containing Perl regex/s that define the tokenization scheme in INPUTFILE file.
If --token is not specified, default token regex file token.regex is searched in the current directory.
--prefix PRE
Specify a prefix to be used for the file names of the cluster files. e.g. If the PRE is the prefix specified then cluster with id=0 will have file name: PRE.cluster.0
If prefix is not specified then prefix is created by concatenating time stamp to the string "expr".
--stop STOPFILE
A file of Perl regexes that define the stop list of words to be excluded from the features.
STOPFILE could be specified with two modes :
AND mode - declared by including '@stop.mode=AND' on the first line of the STOPFILE
OR mode - declared by including '@stop.mode=OR' on the first line of the STOPFILE [Default]
AND mode ignores word pairs in which both words are stop words.
OR mode ignores word pairs in which either word is a stop word.
--remove N
Removes bigrams that occur less than N times.
Default value for this option is 5
--window W
Specifies the window size for bigrams. Pairs of words that co-occur within the specified window from each other (window W allows at most W-2 intervening words) will form the bigram features.
Default window size is 2 which allows only consecutive word pairs.
--stat STAT
Specifies the statistical scores of association. The following are available:
ll - Log Likelihood Ratio [default]
pmi - Point-Wise Mutual Information
tmi - True Mutual Information
x2 - Chi-Squared Test
phi - Phi Coefficient
tscore - T-Score
dice - Dice Coefficient
odds - Odds Ratio
leftFisher - Left Fisher's Test
rightFisher - Right Fisher's Test--rank R
Word pairs ranking below R when arranged in descending order of their test scores are ignored.
Default value for this option is 10
--newLine
If turned on, word pair selection process will not span across newlines.
By default this option is turned off, that is, word pair selection spans across lines.
Other Options :
--help
Displays the quick summary of program options.
--version
Displays the version information.
--verbose
Displays to STDERR the current program status.
OUTPUT
- 1. Cluster ids followed by the assigned labels are directed to STDOUT:
-
Cluster 0 (Descriptive): Bill Clinton, Mariana Islands, Northern Mariana, Pacific island, World Cup, per hour Cluster 0 (Discriminating): Mariana Islands, Northern Mariana, Pacific island, World Cup, per hour Cluster 2 (Descriptive): Bill Clinton, Erik wrote, Inc Within, Jersey And, Lyle Menendez Cluster 2 (Discriminating): Erik wrote, Inc Within, Jersey And, Lyle Menendez Cluster 1: Cluster 3: Cluster -1 (Descriptive): York Times, Undated _ Cluster -1 (Discriminating): York Times, Undated _ - 2. Cluster files, named with the specified prefix or the generated prefix.
SYSTEM REQUIREMENTS
Input to this program should be created by format_clusters.pl
BUGS
AUTHOR
Ted Pedersen, University of Minnesota, Duluth
Anagha Kulkarni, University of Minnesota, Duluth
COPYRIGHT
Copyright (c) 2004-2006,
Anagha Kulkarni, University of Minnesota, Duluth
kulka020@umn.edu
Ted Pedersen, University of Minnesota, Duluth
tpederse@umn.eduThis program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program; if not, write to
The Free Software Foundation, Inc.,
59 Temple Place - Suite 330,
Boston, MA 02111-1307, USA.