pretty_name: GerAV-Twitter
license: other
language:
- de
task_categories:
- text-classification
gated: true
configs:
- config_name: twitter
data_files:
- split: train
path: train-*
- split: validation
path: validation-*
- split: test
path: test-*
extra_gated_heading: Request access to GerAV-Twitter
extra_gated_description: >
GerAV-Twitter is available exclusively for academic research. Please review
the access conditions below before submitting an access request.
extra_gated_button_content: Request access
extra_gated_prompt: >
By submitting this request, you confirm that you have read and agree to the
access conditions described below.
extra_gated_fields:
Full name: text
Institution: text
Institutional email: text
Intended use:
type: select
options:
- Academic research
- Other
Research purpose:
type: text
I agree to the GerAV-Twitter Access Conditions: checkbox
GerAV-Twitter
Dataset Description
GerAV-Twitter is a German authorship verification benchmark belonging to the publication Towards New Heights in German Authorship Verification using Fine-Tuned LLMs on a New Benchmark (Kiefer et al., 2026). The dataset is designed for research on authorship verification and related tasks in computational authorship analysis.
GerAV-Twitter contains authorship verification instances based on
German-language Twitter posts. Each instance consists of two posts,
referred to as post_a and post_b, together with information about
their respective authors and a binary authorship-verification label.
The dataset contains train, validation, and test splits.
The original text of the Twitter posts is not directly distributed with the benchmark. Instead, posts are identified through IDs that can be mapped to the corresponding texts in the original Twitter dataset. Please refer to the paper and the corresponding GerAV GitHub repository for information on obtaining and reconstructing the original texts.
Data Format
| Field | Type | Description |
|---|---|---|
label |
boolean | Whether the two posts were written by the same author |
post_a |
object | Information about the first Twitter post |
post_b |
object | Information about the second Twitter post |
Each post_a and post_b entry contains IDs belonging the the tweet and the author of the tweet.
The authorship verification task is therefore defined as determining
whether post_a and post_b were written by the same author.
Intended Use
GerAV-Twitter is intended for academic and non-commercial research on authorship verification, stylometry, NLP, and related areas.
Access Conditions
You need to satisfy the following conditions for your access request to be accepted:
- The dataset may be used for academic research only. Such use must comply with the applicable Twitter terms of usage and user privacy as stated in the dataset licence of the source dataset.
- Access is restricted to researchers affiliated with a research or educational institution. Please provide an institutional email address with your request.
- The dataset may not be used for commercial purposes or for mass surveillance.
- Access requests are reviewed individually. Additional restrictions or exclusions may be applied based on the information provided in the access request.
- The dataset may not be redistributed or made available to third parties.
By requesting access, you confirm that you have read and agree to these conditions.
Citation
If you use GerAV-Twitter in your research, please cite:
@inproceedings{kiefer-etal-2026-gerav,
title = "{G}er{AV}: Towards New Heights in {G}erman Authorship Verification using Fine-Tuned {LLM}s on a New Benchmark",
author = "Kiefer, Lotta and
Leiter, Christoph and
Takeshita, Sotaro and
Schmidt, Elena and
Eger, Steffen",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.1991/",
doi = "10.18653/v1/2026.findings-acl.1991",
pages = "40050--40069",
ISBN = "979-8-89176-395-1"
}