This repository contains code and data accompanying our ACL 2020 paper, "Unsupervised Domain Clusters in Pretrained Language Models".
The multi-domain German-English parallel data we used in the paper is available here (626MB). It is a new data split we created that avoids duplicate examples and leakage from the train split to the dev/test splits. The original multi-domain data first appeared in Koehn and Knowles (2017) and consists of five datasets available in the Opus website.
Available in a notebook in the src directory. Please contact me in roee.aharoni@gmail.com for any questions.
If you find this useful for your work, please use the following citation:
@inproceedings{aharoni2020unsupervised,
title={Unsupervised domain clusters in pretrained language models},
author={Aharoni, Roee and Goldberg, Yoav},
booktitle = "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
year={2020},
url={https://arxiv.org/abs/2004.02105},
publisher = "Association for Computational Linguistics"
}