CausalScore

a model-based metric for dialogue response evaluation.

Setup

We use conda to manage python packages.

conda env create -n dialog_rel --file environment.yml
conda activate dialog_rel

Usage

annotation/amazon_mturk_interface.html is the annotation interface on amazon mturk.

run_text_classification.py is the training script for unconditional independence classifier.

run_condtional_classifier_EM.py is the training script for conditional independence classifier.

interact_models.py is the script to interact with the trained models.

CGDIALOG+ is in cgdialog+ folder.