-------------------------------
evalC
-------------------------------

evalC computes the accuracy of a bracketed test corpus when compared with a gold version.
The two treebanks should follow the Penn Treebanks formats.
The output of evalC is very similar to the one provided by evalB (http://nlp.cs.nyu.edu/evalb/).
In addition to the overall and per-sentence bracketing accuracy, evalC gives more in depth evaluation about:
- per-category accuracy
- per-category crossing bracket
- category mismatches

USAGE: 
java -jar evalC.jar [-p paramFile] [-log logFile] goldFile testFile outputFile

The directory sample, contains an example pair of gold and test file (sample.gld, sample.tst),
and the output of evalB (sample.evalB) and evalC (sample.evalC) when used with specific parameter files (param.evalB, param.evalC).
IMPORTANT: the two parameter files are not equivalent and cannot be used interchangeably.

-------------------------------
evalC_GUI
-------------------------------

evalC_GUI is the graphical version of evalC.
It displays each gold/test structure and marks in red the non-matching constituents (in red) and incorrect pos-tags (in green). 
From the File menu it is possible to:
- Load a Gold File (should be in Penn Treebanks format)
- Load a Test File (should be in Penn Treebanks format)
- Load a Parameter File (i.e. sample/evalC.prm)
- Compare (display the Gold/Test treebanks and mark the differences)
- Compare with output (display the Gold/Test treebanks with differences and in addition output an evaluation file.
- Export to PDS (export the current Gold/Test pair of structures to a PDF document)
- Export ALL to PDS (export the Gold/Test pairs of structures of the two treebanks to a PDF document, one pair per page)

USAGE: 
java [-Xmx100M] -jar evalC_GUI.jar

--------------------------------------------------------------
For questions/comments contact me: f.sangati (at) uva (dot) nl
--------------------------------------------------------------

