Code for the Paper "An Item Response Framework for Persuasion". The paper can currently be found on arxiv:
The irt_lib
folder contains shared code for models and data preparation. The debates
and editorials
folders contain dataset specific processing, results and analysis. In each subfolder, the README file contains overall details on where to download the raw data, then the DataAssembly.ipynb
notebooks show how the data was transformed into feature+label pairs. Finally, the Models.ipynb
notebooks contain the experiments included in the paper.
The training was conducted using Amazon SageMaker on a ml.p3.2xlarge
machine. The model_env.lst
contains the python requirements/library versions used during training. The data_env.lst
file contains the python requirements for preprocessing.
Note: For the "SpaCy" preprocessing, you will need to download the models using python -m spacy download en_core_web_md