StyLEx: Explaining Style Using Human Lexical Annotations

This repository provides datasets and code for preprocessing, training and testing models for style classification with human lexical annotations with the official Hugging Face implementation of the following paper:

StyLEx: Explaining Style Using Human Lexical Annotations
Shirley A. Hayati, Kyumin Park, Dheeraj Rajagopal, Lyle Ungar, Dongyeop Kang
EACL 2023

Installation

The following command installs all necessary packages:

pip install -r requirements.txt

The project was tested using Python 3.8.

Models

Model Checkpoints

Model	Style	F1 (Orig)	F1 (Hummingbird)	F1 (OOD)
BERT	Politeness	0.96	0.91	0.87
BERT	Sentiment	0.67	0.91	0.75
BERT	Joy	0.88	0.92	0.73
BERT	Sadness	0.89	0.94	0.78
BERT	Fear	0.96	0.92	0.80
BERT	Disgust	0.86	0.81	0.74
BERT	Anger	0.89	0.82	0.78
BERT	Offensiveness	0.97	0.87	0.88

Citation

If you find this work useful for your research, please cite our papers:

@article{hayati2022stylex,
  title={StyLEx: Explaining Styles with Lexicon-Based Human Perception},
  author={Hayati, Shirley Anugrah and Park, Kyumin and Rajagopal, Dheeraj and Ungar, Lyle and Kang, Dongyeop},
  journal={arXiv preprint arXiv:2210.07469},
  year={2022}
}