Authorship-Deobfuscation

This repository contains code and data used in the paper "A Girl Has A Name, And It’s ... Adversarial Authorship Attribution for Deobfuscation" accepted at ACL 2022.

Requirements

1. Install Python3
2. Download trained models and data for training and testing in the main folder.
3. Install libraries needed for the project through

  pip3 install -r requirements.txt

Usage

Evaluate prediction accuracy for obfuscated texts

To evaluate directly using previous data, run the following in the main folder.

  python3 eval.py

Evaluate performance on METEOR score

To evaluate the METEOR score from previous data, run the following in the main folder.

  python3 CalcMeteor.py