This is the repository for our paper, Maintaining Common Ground in Dynamic Environments (Udagawa and Aizawa, TACL 2021).
Install and activate pyenv.
pyenv install 3.7.7
pyenv local 3.7.7
Install and activate poetry.
poetry install
poetry shell
You also need to unzip and place the pregenerated scenarios (environments) used in our dataset: please refer to the scenarios directory for details.
To generate new scenarios (environments) for Dynamic-OneCommon, you can use the scenarios directory.
To collect and review human dialogues (e.g. from Amazon Mechanical Turk), you can use the webapp directory.
To analyze the collected dataset, you can use the data directory.
To train and evaluate baseline dialogue models, you can use the experiments directory.