This repository contains code for the paper "It's Morphin' Time! Combating Linguistic Discrimination with Inflectional Perturbations" (to be presented at ACL 2020).
Authors: Samson Tan, Shafiq Joty, Min-Yen Kan, and Richard Socher
To generate adversarial examples for one of the implemented models, run the corresponding run_morpheus_*
script.
Morpheus can be easily implemented for a custom task, dataset, or model by following the structure of existing classes:
MorpheusBase
implements the methods common to all Morpheus implementations; Morpheus<Task>
implements methods common to a specific task/dataset, Morpheus<Model><Task>
implements methods specific to a particular model (usually the init
and morph
methods).
Coming soon.
@inproceedings{morpheus20,
author = {Samson Tan and Shafiq Joty and Min-Yen Kan and Richard Socher},
title = {It’s {M}orphin’ {T}ime! {C}ombating Linguistic Discrimination with Inflectional Perturbations},
booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
year = {2020},
address = {Seattle, Washington},
numpages = {9},
publisher = {Association for Computational Linguistics},
url = {https://arxiv.org/abs/2005.04364}
}