Support for users of CCG software closed 7-27-20

["C. Howell" <caden AT cs.wisc.edu>]

Hello!

I am on a team planning to use LBJCoref.jar as part of a larger research project (the MR FAUST team, where we are UIUC's teammates.)  All of the other parts of the project (written in Java) run on both Linux and Windows, and we do most of our debugging and development on Windows machines.  Unfortunately, it will cost us many hours moving our debugging and development tools back and forth between Linux and Windows if we cannot get the LBJCoref.jar to run on Windows.

In order to isolate the issue we are having, I attempted to run the jar on Linux and on Windows.  On Linux (RedHat) it works.  Here is an example with a small file containing a sports story.

-bash-3.2$ java -classpath .:LBJ2.jar:LBJ2Library.jar:LBJPOS.jar -jar LBJCoref.jar AFP_ENG_small.sgm 
< *DOC*_7 id = " *AFP_ENG_19950107.0244*_8 " type = " story " > < HEADLINE > *49ers*_3 rip Bears , reach *NFL*_10 semi-finals < / HEADLINE > < DATELINE > *SAN FRANCISCO*_5 , *California*_2 , Jan 7 ( *AFP*_8 ) < / DATELINE > < TEXT > < *P >*_9 *William Floyd*_11 rushed for three touchdowns and *Steve Young*_12 scored two more , moving the *San Francisco 49ers*_3 one victory from the Super Bowl with a 44-15 *American*_0 football rout of *Chicago. < / P >*_9 < *P*_9 > By winning the *National Football League*_4 ( *NFL*_10 ) playoff game , *the 49ers*_3 will host *the winner of *Sunday 's Dallas-Green Bay*_6 game on January 15*_1 to decide a berth in the January 29 championship game at *Miami. < / P >*_9 < / TEXT > < / *DOC*_7 >

On Windows, it does not work.  I changed the classpath appropriately.  Here is the error:

C:\Workspaces\MachineReading\scratch\jars>java -classpath .;LBJ2.jar;LBJ2Library.jar;LBJPOS.jar -jar LBJCoref.jar AFP_ENG_small.sgm 
Exception in thread "main" java.lang.RuntimeException: java.io.FileNotFoundException: java.io.FileNotFoundException
        at edu.brandeis.cs.steele.wn.FileBackedDictionary.lookupIndexWord(FileBackedDictionary.java:180)
        at edu.illinois.cs.cogcomp.lbj.coref.features.WordNetTools.lookupIndexWordSafe(WordNetTools.java:586)
        at edu.illinois.cs.cogcomp.lbj.coref.features.WordNetTools.getIndexWord(WordNetTools.java:512)
        at edu.illinois.cs.cogcomp.lbj.coref.features.WordNetTools.getIndexNoun(WordNetTools.java:492)
        at edu.illinois.cs.cogcomp.lbj.coref.features.WordNetTools.getHypernymStrings(WordNetTools.java:389)
        at edu.illinois.cs.cogcomp.lbj.coref.learned.bioHypernyms.classify(bioHypernyms.java:46)
        at edu.illinois.cs.cogcomp.lbj.coref.learned.MentionDetectorMyBIOHead$$1.classify(MentionDetectorMyBIOHead$$1.ja
va:143)
        at LBJ2.learn.Learner.getExampleArray(Learner.java:358)
        at edu.illinois.cs.cogcomp.lbj.coref.learned.MentionDetectorMyBIOHead.getExampleArray(MentionDetectorMyBIOHead.j
ava:591)
        at LBJ2.learn.Learner.classify(Learner.java:203)
        at edu.illinois.cs.cogcomp.lbj.coref.learned.MentionDetectorMyBIOHead.classify(MentionDetectorMyBIOHead.java:214
)
        at edu.illinois.cs.cogcomp.lbj.coref.learned.MentionDetectorMyBIOHead.classify(MentionDetectorMyBIOHead.java:210
)
        at edu.illinois.cs.cogcomp.lbj.coref.learned.MentionDetectorMyBIOHead.discreteValue(MentionDetectorMyBIOHead.jav
a:262)
        at edu.illinois.cs.cogcomp.lbj.coref.decoders.BIODecoder.decode(BIODecoder.java:50)
        at edu.illinois.cs.cogcomp.lbj.coref.decoders.ExtendHeadsDecoder.decode(ExtendHeadsDecoder.java:58)
        at edu.illinois.cs.cogcomp.lbj.coref.io.loaders.DocLoader.getPredMents(DocLoader.java:256)
        at edu.illinois.cs.cogcomp.lbj.coref.io.loaders.DocLoader.loadDoc(DocLoader.java:178)
        at edu.illinois.cs.cogcomp.lbj.coref.CorefPlainText.main(CorefPlainText.java:88)

My guess is that this has something to do with the case insensitivity of Windows.  My first thought was to download the source code for LBJCoref.jar and run it in the debugger.  Possibly I could make a change and recompile it.  However, it appears to expect a copy of the  ACE 2004 corpus.  I contacted LDC and discovered that unfortunately, we are not a member of LDC and we are not budgeted the $3000 required to use this corpus.   

I can see that the error is occurring in edu.brandeis.cs.steele.wn.FileBackedDictionary.lookupIndexWord, but I wonder if the error is caused by the way it is called.  For example, if the filename passed to FileBackedDictionary were lowercase and all of the dictionary files on disk were lowercase, would it eliminate the error?

At this point I'm open to any and all suggestions.  My team is nearly resigned to working around this problem, but I can see it is causing us to waste time.

Thanks for any help,

Caden Howell
caden AT cs.wisc.edu
University of Wisconsin - Madison 
Department of Computer Science