sjmeis/blogtext_privacy

Blog Text Dataset

This dataset contains blog posts collected from blogger.com, split into two curated configurations: author10 and topic10.

The primary purpose of this dataset is to enable the evalaution of text privatization methods in protecting sensitive attributes (authorship, age, gender, profession).

Dataset Structure

The dataset is split into two distinct subsets/configurations:

1. author10

Consists of 15,070 blog posts belonging to exactly 10 unique authors. It is ideal for author identification and stylometric profiling tasks, including for designing privacy-preserving mechanisms in defense.

• id: Unique identifier for the author (integer from 1 to 10).
• topic: The topic category designated by the blogger, i.e., their profession (includes 5 unique domains: Technology, Non-Profit, Arts, Fashion, Internet).
• text: The raw text content of the blog post.

2. topic10

Consists of 14,259 blog posts across exactly 10 topics / professions. It includes demographic labels, making it suitable for atribiute protection, multi-label text classification, and demographic bias analysis.

• gender: Gender of the author (male or female).
• age: Age bracket of the author (13-22, 23-24, 25-26, 27-33, 34-48).
• topic: One of the 10 topic catrgories of the blogger's self-reported profession (Technology, Arts, Education, Communications-Media, Internet, Non-Profit, Engineering, Law, Science, Government).
• text: The raw text content of the blog post.

Usage

You can load either subset using the Hugging Face datasets library:

from datasets import load_dataset

# Load the author10 subset
dataset_author = load_dataset("sjmeis/blogtext_privacy", "author10")

# Load the topic10 subset
dataset_topic = load_dataset("sjmeis/blogtext_privacy", "topic10")

Citation

Please consider citing our work which introduced these subsets:

@inproceedings{meisenbacher-matthes-2024-thinking,
    title = "Thinking Outside of the Differential Privacy Box: A Case Study in Text Privatization with Language Model Prompting",
    author = "Meisenbacher, Stephen  and
      Matthes, Florian",
    editor = "Al-Onaizan, Yaser  and
      Bansal, Mohit  and
      Chen, Yun-Nung",
    booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.emnlp-main.324/",
    doi = "10.18653/v1/2024.emnlp-main.324",
    pages = "5656--5665"
}

In addition, please make sure to cite the original work making this full dataset public:

@inproceedings{schler2006effects,
  title={Effects of age and gender on blogging.},
  author={Schler, Jonathan and Koppel, Moshe and Argamon, Shlomo},
  booktitle={AAAI spring symposium: Computational approaches to analyzing weblogs},
  volume={6},
  pages={199-205},
  year={2006},
  url={https://aaai.org/papers/0039-ss06-03-039-effects-of-age-and-gender-on-blogging/}
}