Datasets:
language:
- de
task_categories:
- text-classification
size_categories:
- 1K<n<10K
multilinguality:
- monolingual
configs:
- config_name: default
data_files:
- split: train
path: train.parquet
- split: test
path: test.parquet
Information
This dataset shows 1785 manually annotated tweets from German politicians during the election year 2021 (01.01.2021 - 31.12.2021). The tweets were annotated by 6 academics which were separated into two different groups. So every group of 3 people annotated the sentiment of ~900 tweets. For every tweet, the majority label was built. The annotation result had a moderate Kappa agreement.
Preprocessing
The source for this version of the dataset is located here.
For better processing the line breaks of the texts are removed.
The numbers for answers, retweets and favorites are removed from the text and also the phrase "Diesen Thread anzeigen" (Show this thread). Both aren't part of the tweet and were most likely added by the crawling tool.
The preprocessing steps can be reproduced with the cleaner.py script.
Annotation
The tweets were annotated as follows:
- 1 if the sentiment of the tweet is positive
- 2 if the sentiment of the tweet is negative
- 3 if the sentiment of the tweet is neutral
Citation
@inproceedings{schmidt-etal-2022-sentiment,
title = "Sentiment Analysis on {T}witter for the Major {G}erman Parties during the 2021 {G}erman Federal Election",
author = "Schmidt, Thomas and
Fehle, Jakob and
Weissenbacher, Maximilian and
Richter, Jonathan and
Gottschalk, Philipp and
Wolff, Christian",
editor = "Schaefer, Robin and
Bai, Xiaoyu and
Stede, Manfred and
Zesch, Torsten",
booktitle = "Proceedings of the 18th Conference on Natural Language Processing (KONVENS 2022)",
month = "12--15 " # sep,
year = "2022",
address = "Potsdam, Germany",
publisher = "KONVENS 2022 Organizers",
url = "https://aclanthology.org/2022.konvens-1.9",
pages = "74--87",
}