Enron28

This dataset contains a curated selection of sent emails from the historical Enron email corpus. It is designed specifically for text privatization, author identification, stylometry profiling, and text classification tasks, focusing on the most prolific writers in the corpus.

Dataset Structure

The dataset consists of 12,283 rows:

• author: The identifier of the email sender (e.g., dasovich-j, germany-c). There are 28 unique authors in total, heavily represented by the top writers.
• text: The raw text content extracted from the body of the sent email.

Usage

You can easily load this dataset into your environment using the Hugging Face datasets library:

from datasets import load_dataset

# Load the dataset from your Hugging Face repository
dataset = load_dataset("sjmeis/enron28")

Citation

Please consider citing the following work which prepared the split presented here:

@inproceedings{meisenbacher-etal-2025-impact,
    title = "On the Impact of Noise in Differentially Private Text Rewriting",
    author = "Meisenbacher, Stephen  and
      Chevli, Maulik  and
      Matthes, Florian",
    editor = "Chiruzzo, Luis  and
      Ritter, Alan  and
      Wang, Lu",
    booktitle = "Findings of the Association for Computational Linguistics: NAACL 2025",
    month = apr,
    year = "2025",
    address = "Albuquerque, New Mexico",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.findings-naacl.32/",
    doi = "10.18653/v1/2025.findings-naacl.32",
    pages = "514--532",
    ISBN = "979-8-89176-195-7"
}

In addition, please also reference the original work presenting the Enron corpus:

@inproceedings{klimt2004introducing,
  title={Introducing the Enron corpus.},
  author={Klimt, Bryan and Yang, Yiming},
  booktitle={CEAS},
  volume={45},
  pages={92--96},
  year={2004}
}