EOPRunner Command Line Parameters

EOPRunner is a command line standalone Java class serving as a unique entry point to the EOP main included functionalities. The class that is located in the util package can call both the linguistic analysis pipeline to pre-process the data to be annotated and the selected entailment algorithm (EDA).

Set-up

Except for the situation when we want only scoring of an existing results file, some of the parameters in this section are necessary.

• -config <Configuration file for the EDA>

Almost all parameters can be passed to the EOPRunner through the configuration file: the language, the lap, the eda, the trainDir and testDir (directories with xmi files for training and testing), the model for training. The values provided through the command line have precedence over the values specified in the configuration file.

• -eda <The class name of the EDA to be used>
• -language <Language of the dataset (DE|EN|IT)>
• -lap <The LAP to be used>

Training

• -train (If this option is present, perform training)
• -trainFile <Input file in RTE format with TE pairs for training>
• -trainDir <Directory for LAP-produced xmi files from training data>

If -trainFile and -trainDir are present without the -train option, then the EOPRunner will preprocess the trainFile and write the xmi files to trainDir (with the LAP specified in the configuration file, or the lap and language given through the command line). If the

Testing

• -test (If this option is present, perform testing)
• -testFile <Input file in RTE format with TE pairs for testing>
• -testDir <Directory for LAP-produced xmi files for test data>

As for training, if -testFile and -testDir are present without the -test option, then the EOPRunner will preprocess the testFile and write the xmi files to testDir (with the LAP specified in the configuration file, or the lap and language given through the command line).

Instead of providing a test file, we can provide a text/hypothesis pair (the -test option must also be given)

• -text <The text part of a text/hypothesis pair>
• -hypothesis <The hypothesis part of a text/hypothesis pair>

Model

To use a specific model for training and testing, the best way is to update the configuration file. Otherwise give in on the command line, and the EOPRunner will change the configuration file accordingly.

• -model <Input model file (obtained through previous training)>

Output and Results

If no results file is given, the name of the results file will be composed from the name of the output directory (default "./"), and the name of the configuration file.

• -results <The name of the (xml formatted) results file>
• -output <Directory for the run's output>

If the -score option is present, it scores the given results file (accuracy and precision/recall/f1-score)

• -score (If present, it will compute the scores on the test data or the results file)