cooccurrence-lexicon
You must have the BBN named entity annotations for the Penn Treebank, and its parallel pronoun resolution file, in the WSJ folder under the root directory. This means that relative to the folder that contains clusters.sh and clusters.bat, the files WSJ/WSJ.pron, WSJ/wsj00a.qa.bj.lm, etc. should exist.
If you do not have access to the BBN NE corpus, I can provide it on request as long as it is strictly for an educational purpose or otherwise satisfies Fair Use.
Executing clusters.sh or clusters.bat will go through the full process. It will generate the files clusters.gephi (a Gephi project file), clusters.gexf, clusters.pdf (visualization, nodes are labeled with IDs), and clusters.xml (the key that maps node IDs in the visualization to entity names) in the out folder.
If comentions.xml and aliases.xml already exist in the out folder, they will be used. Otherwise, all of the intermediary files will be generated from scratch using the BBN corpus and its pronoun resolution file. NOTE: if you are regenerating the intermediary files, the alias resolution program will take a long time to complete. Maybe even up to half an hour! The graph layout algorithms may take a minute or two to complete too. NOTE: I had file permission problems when Dropbox was running. If you have a file sync service running, you may have to temporarily disable it while the corpus cleaner program is running.
I generated charts and tables using the included statistics.R script, which requires that the out/network.xml file be generated. This file can be generated with comentions.sh or comentions.bat.