Code and dataset for our EMNLP 2022 paper:
AEG: Argumentative Essay Generation via A Dual-Decoder Model with Content Planning
python==3.7.11
datasets==2.5.2
nltk==3.6.5
numpy==1.21.6
torch==1.9.0+cu111
transformers==4.14.1
python get_tfidf_keywords.py
python preprocess.py
bash run.sh
See the results in ./outputs
.
For the pre-training results, please first train the model with the CNN-DailyMail dataset with the same procedure.