README

The purpose of CombiTagger is to read files generated by individual PoS taggers and use them to develop and evaluate combined taggers according to a given combination algorithm.

Compilation
-----------
To compile this project you need the swt.jar  (http://www.eclipse.org/swt/) for your platform.

swt.jar can, for example, be downloaded from: 
http://download.eclipse.org/eclipse/downloads/
(select the latest release and then search for "SWT")

Copy the swt.jar for your platform into the lib directory and the run ant (to build the CombiTagger.jar file).
The swt files are copied into the CombiTagger.jar file during the build (see build.xml)

Testing
-------
The exampleFiles directory contains three tagger output files and a gold standard file. These files can be used for testing the functionality of CombiTagger.

Running
-------
If no command line parameters are given, CombiTagger starts with a GUI.
Possible command line parameters can be seen with: java -jar dist/CombiTagger.jar -help

For example, to run CombiTagger using simple voting on three tagger files and writing the output to a file:
java -jar dist/CombiTagger.jar -t exampleFiles/tagger1.txt exampleFiles/tagger2.txt exampleFiles/tagger3.txt -o exampleFiles/results.txt

Documentation
-------------
Verena Henrich, Timo Reuter and Hrafn Loftsson. CombiTagger: A System for Developing Combined Taggers. In Proceedings of the 22nd International FLAIRS Conference, Special Track: "Applied Natural Language Processing". Sanibel Island, Florida, USA. © 2009 AAAI.  

The paper is accessible at: http://www.ru.is/~hrafn