The XAR system is implemented in Java. The source code under 'src' however needs to be linked with a number of 'jar' files, given that XAR uses (i) GATE and (ii) TuProlog for underlying tokenization and deductive inference capabilities respectively .

Software (Prerequsities):

You will need to have installed (i) Java JDK (I have used v 1.5.0) and a JRE (I have used v 1.6.0), (ii) GATE, this is the underlying tokenizer and name-entity recognizer used, download GATE (v 3.1, I have not tested with GATE 4.0) from here and install it, (iii) Eclipse, this is just the IDE I used. You should of course be able to compile XAR without Eclipse as well.

Environment variables:

Set JAVA_HOME should be set to the path of your (Java) JDK.

Include the path to <JRE-ROOT>/lib in your classpath (<JRE-ROOT> is the root directory for your JRE)

Set (the environment variable) gate.home to <GATE-ROOT> (where <GATE-ROOT> is the root directory of where GATE gets installed on your machine)

Compilation Steps: (Eclipse)

Note: I have succesfully compiled using the above JDK and JRE versions, in Eclipse 3.1 on a Windows XP machine.

0) Create a new (Java) project in the Eclipse environment.

1) Import the files in the src directory (Right click on the new project created in Eclipse, select "import", and then the src directory.

2) Link to the required jar files. Right click on the package (default) created, select "Build Path", then "Configure build path", then under 'Libraries' select "Add external jars" and:

Select gate.jar under <GATE-ROOT>/bin/. Similarly link to 'tools.jar' which should be under your <JDK-ROOT>/bin/. Finally also link to all the jar files in the 'jars' directory provided in the XAR distribution.

The src files should all compile without any errors. The main executable is 'Extractor' which is what you invoke to run XAR.

Note: This application requires a large(r) amount of virtual memory. Set it to -Xmx1024m under 'Arguments' (VM) in the 'Run' window in Eclipse.

Invoking 'Extractor' should bring up the main XAR menu.

NOTE: It is recommended that you first try running some pre-assembled applications provided, and also assemble at least one application on your own with this installation. Then you can integrate in the deeper analyzers as described below.

Additional: StanfordParser Integration

1. Download the StanfordParser source code available here.

2. In the new project that you created for XAR, now also import the file system at /edu/src.

3. In XAR.java do a search for the comment label //LABEL-SNP. Decomment the code at these locations (there are 5 such instances)

That's it. The StanfordParser is integrated in, and you will be able to use it for deep feature extraction.