Project Setup & Requirements

###Project Setup & Requirements

There are no special requirements for this project. All necessary files are contained in the project itself. It can be run after importing the project in a Maven compatible IDE (e.g. Eclipse for Java Developers ) and selecting an experiment to run.

####Project Structure
The project has two distinct parts. ../src/main/ contains all packages, classes and resources necessary to run experiments. (Experiments are classes, doing the actual Complex Word Identification.) ../src/test/ contains all unit test and various other testing classes for this project.

###Run Configurations

Experiments are located in the project folder under ../src/main/java/de/unidue/langtech/pp/experiments. There are currently 5 preconfigured experiments available:

• Complexity_ZeroR_CV: Uses the Weka Engine's ZeroR (majority class) Classifier in a cross validation experiment.
• Complexity_Bayesline_CV: Uses the Weka Engine's Naive Bayes Classifier in a cross validation experiment.
• Complexity_J48_CV: Uses the Weka Engine's J48 (C4.5) Decision Tree Classifier in a cross validation experiment.
• Complexity_RandomForest_CV: Uses the Weka Engine's RandomForest Classifier in a cross validation experiment.
• Complexity_Comparison_CV: Uses all of the aforementioned Weka Classifiers (except ZeroR) in a cross validation experiment.

The number of folds for these is set to 10 by default and can be adjusted, by changing the NUM_FOLDS value in the java file. The CORPUS_FILEPATH_TRAIN is set to CORPUS_SMALL by default and can be changed to CORPUS_LARGE.
For more advanced customisation options, please refer to the DKPro TC User Guide.

###Output

Output files are stored in the project folder under ../src/main/resources/output/. Each experiment has its own subfolder and generates a report for every run, located in subfolders, named by corresponding timestamps of the run.

######Label Coding

• 1: Complex Word
• 0: Non-Complex Word