Author: Dennis Aumiller
- OS: Tested on Ubuntu 20.04
- Python3.7 or later
- 7 GB disk memory for datasets
- CUDA-compatible GPU preferable for reproduction of inference runs
Note: This has been tested on Ubuntu 20.04 LTS. A successful reproduction on other systems is not guaranteed!
Furthermore, this assumes you have Python3.7+ installed on your system.
To reproduce the content necessary for the paper's experiments,
simply install the library according to the README.md
's instructions:
First, install the necessary dependencies with
python3 setup.py install
and then set up the library with pip
:
python3 -m pip install .
Also ensure that you have the unzip
shell command installed:
sudo apt-get install unzip
The master script produces the majority of artifacts required to reproduce this work. To execute it, navigate to the respective experimental folder and run these two commands:
cd experiments/german_summarization_datasets
./master_script.sh
This will automatically download all the necessary data for the experiments (careful, this is close to 7 GB) and execute the individual experiments.
Also see the comments in master_script.sh
for more information on the respective commands and tools.