Repository for the paper "Long-Tail Zero and Few-Shot Learning via Contrastive Pretraining on and for Small Data", Nils Rethmeier, Isabelle Augenstein, 2020

To be updated soon. In the meantime, you can aquire code and data by reaching out to the paper email.

CLESS, uses contrastive label-embedding self-supervision to enable data efficient text encoder pretraining that is inherently zero-shot transferable and improves long-tail minority learning.

We examine learning on a challenging long-tailed, low-resource, multi-label text classification dataset with noisy, highly sparse labels and many minority concepts.

This is an updated version and the future code repository for the paper "Long-Tail Zero and Few-Shot Learning via Contrastive Pretraining on and for Small Data", Nils Rethmeier, Isabelle Augenstein, 2 Oct 2020, https://arxiv.org/abs/2010.01061 -->

@misc{rethmeier2020longtail,
    title={Long-Tail Zero and Few-Shot Learning via Contrastive Pretraining on and for Small Data},
    author={Nils Rethmeier and Isabelle Augenstein},
    year={2020},
    eprint={2010.01061},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}