German Politicians' 2021 Election Tweets with Sentiment Annotations

Featuring 1,785 manually annotated tweets from German politicians during the 2021 election year. The sentiment labels were assigned by two groups of three academics each, with a moderate Kappa agreement score for the annotations.

Use Cases

• Analyze sentiment label distributions across tweets from different politicians during the 2021 election period.
• Train a text classifier to predict sentiment labels on German political tweets.
• Study the relationship between tweet content and the assigned sentiment label from the academic annotators.

Strengths

• 1,785 manually annotated tweets provide a labeled corpus for analysis.
• Annotations were performed by six academics, providing a multi-rater perspective.
• Data is temporally focused on the 2021 German election year, offering contextual relevance.

Limitations

• The dataset size of 1,785 tweets is relatively small for training complex machine learning models.
• Annotation agreement is described as only 'moderate', indicating potential label noise.
• The dataset is limited to a single year and geographic context, reducing generalizability.

Provenance

Source

huggingface, author 'mox'.

Collection Method

Tweets were manually annotated for sentiment by six academics, with a majority label derived from two groups of three.

Time Range

01.01.2021 - 31.12.2021

Freshness

Data covers the year 2021; the dataset record was last updated in May 2022.

Geography

Germany