Twitter Hate Speech Dataset with 242,738 Tweets

242,738 rows of uncleaned Twitter data combine three established hate speech datasets. The collection has only two columns, 'tweet' and 'labels'. It was created by TLeonidas and last updated on April 6, 2024.

Use Cases

• Train hate speech classifiers based on the 'labels' column.
• Benchmark model performance against a large, aggregated corpus of social media text.
• Analyze linguistic patterns in abusive online language based on the 'tweet' content.

Strengths

• Aggregates 242,738 data points from three established source datasets.
• Combines multiple sources, which may provide broader coverage of hate speech phenomena.

Limitations

• Data is described as 'uncleaned', which likely requires significant preprocessing.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2024-04-06 00:03:47; freshness should be verified.

Provenance

Source

Combination of tdavidson/hate_speech_offensive, LennardZuendorf/Dynamically-Generated-Hate-Speech-Dataset, and ucberkeley-dlab/measuring-hate-speech from Hugging Face.

Collection Method

Aggregation of existing datasets.

Freshness

Last updated 2024-04-06 00:03:47.