These are the scripts used for training, development and testing of the experiments for predicting the 4 HTER components jointly with a single layer perceptron. Requirements ============ Python 2 requirements can be installed by creating a python virtual environment and then running pip install -r requirements.txt Simple overview of scripts ========================== ter.py: rerun tercom in order to generate the 4 HTER edits check_hter: check the regenerated hter against the hter provided by the organizers train_test.py: create a model and use it to measure the performance on a development set predict.py: apply an existing model to make new predictions confidence.py: run a significance test between the golden labels and two lists of predictions. the augmented feature set was produced with Qualitative. the scripts will be uploaded to a repository, after adding some more documentations and inline comments.