Enriched variant of DeepNLG, introducing information flow hints.
It is recommended to run the code in a docker container or a designated conda envronment.
- clone the repository.
- install requirements with
pip install -r requirements.txt
. (possibly changetorch
version depending on your version ofcuda
)
The code is compatible with cuda
, and it is recommended to run it on a cuda
machine, but runs on cpu as well.
Enriched dataset is located in data
folder by default.
Running DatasetGenerator.py
will create the enriched dataset from DeepNLG located in DeepNLG_data
To generate text from FlowNLG and measure BLEU metric with respect to references run EvalOracle.py