Datasets:

Modalities:
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
book_title
stringlengths
1
455
original_publication_year
stringclasses
210 values
text
stringlengths
270
28.5M
gutenberg_release_date
stringlengths
10
10
Familiar Studies in Homer
null
"\n\n\n\n FAMILIAR STUDIES IN HOMER\n\n\n\n\n\n\n-----------------------------(...TRUNCATED)
04/05/2021
Ecclesiastical History of England from the Opening of the Long
null
"FROM THE OPENING OF THE LONG PARLIAMENT TO THE DEATH OF OLIVER CROMWELL ***\n\n\n\n\n (...TRUNCATED)
01/06/2021
Out of Mind
null
"\n\n\nProduced by Greg Weeks, Mary Meehan and the Online\nDistributed Proofreading Team at http://w(...TRUNCATED)
12/30/2019
My Wayward Pardner; or, My Trials with Josiah, America, the Widow Bump, and Etcetery
null
"\n\n\nProduced by KD Weeks, hekula03 and the Online Distributed\nProofreading Team at https://www.p(...TRUNCATED)
11/23/2020
Southey
null
"\n\n\n\nProduced by Juliet Sutherland and the Online Distributed\nProofreading Team at https://www.(...TRUNCATED)
04/30/2020
Cattle
null
"\n\n\nProduced by Mary Glenn Krause, Martin Pettit and the Online\nDistributed Proofreading Team at(...TRUNCATED)
08/12/2020
The Footpath Way: An Anthology for Walkers
null
"\n\n\nThe Footpath Way\n\nAN ANTHOLOGY FOR WALKERS\n\nWITH AN INTRODUCTION BY\n\nHilaire Belloc\n\n(...TRUNCATED)
06/25/2019
Eight Million Dollars From Mars!
1954
"\n\n\n\n\n Eight Million Dollars From Mars!\n\n By Wins(...TRUNCATED)
10/01/2021
The Disembodied Man
1954
"\n\n\n\n The Disembodied Man\n\n By Larry Maddoc(...TRUNCATED)
09/13/2021
Medical Inquiries and Observations, Vol. 1
null
"\n\nE-text prepared by MWS, Jens Nordmann, Bryan Ness, and the Online\nDistributed Proofreading Te(...TRUNCATED)
02/26/2019

Books from Project Gutenberg (after PG-19)

This dataset contains 9,542 books collected from Project Gutenberg, an online library for free e-books.

Specifically, we collect books that have been added to Project Gutenberg after the last book that has been included in the widely used PG-19 dataset. Of all books included in PG-19, the latest release date on Project Gutenberg was February 10, 2019. We use an open source library to download all English books that were added to Project Gutenberg after this date (and adapt the code here). As preprocessing, we only consider the text between the explicit start and end of the uniformly formatted text files.

This data has been collected as part of the experimental setup of the paper "Did the Neurons Read your Book? Document-level Membership Inference for Large Language Models" (link). The goal was to create a dataset containing representative non-member documents compared to PG-19 to develop and evaluate a Membership Inference Attack (MIA) against a Large Language Model (LLM) trained on data containing PG-19.

We here release the data we have used to generate the results discussed in the paper, mainly to facilitate further research in similar directions. Importantly, research beyond the study in the paper (here) suggests that this dataset exhibits a serious distribution shift in language compared to books in PG19. Hence, it is not recommended to use this data -at least not in its current form- as non-member data to develop and evaluate post-hoc MIAs against LLMs.

Of course, the dataset also represents a rich source of natural language from the literature -most of which should be in the public domain in the US- and could also be used for other purposes.

If you found this dataset helpful for your work, kindly cite us as

@article{meeus2023did,
  title={Did the neurons read your book? document-level membership inference for large language models},
  author={Meeus, Matthieu and Jain, Shubham and Rei, Marek and de Montjoye, Yves-Alexandre},
  journal={arXiv preprint arXiv:2310.15007},
  year={2023}
}
Downloads last month
63