* Introduction
This program implemented a sentence compression parser, based on the method proposed by Ryan McDonald.

* Program Organization
** CompressionParser/
    data/   : store training data, testing data (NEU Corpus) and other files.
    eval/   : store evaluation data for mteval-v11b.pl script.
    script/ : evaluate script and preprocess corpus data.
    src/    : source code
    target/ : store maven compiled files
    pom.xml : maven config file
    Readme

* Dependency Libraries
    This program depends on "trove.jar", and we use 2.0.2 version.

* Compile Code
We use maven organize our code, you simply change to the project directory  and then type:
    mvn package

After above process, you will get a SCParser-1.0.jar in the "target" directory.

* Usage Example
Following process, we assume you are in the CompressionParser directory.

1. Train
java -cp target/SCParser-1.0.jar:target/dependency/trove4j-2.0.2.jar CompressionParser -train -t 10 -k 10

2. Test
java -cp target/SCParser-1.0.jar:target/dependency/trove4j-2.0.2.jar CompressionParser -test -k 10

* Reference
[1]. Discriminative Sentence Compression with Soft Syntactic Evidence. Ryan McDonald.