Hate Speech and Offensive Language Dataset from Twitter

The 'hate-speech-and-offensive-language' dataset, introduced by Davidson et al. in 2017, classifies tweets into three categories: hate speech, offensive language, or neither. The dataset was created by researchers from Cornell University's Department of Sociology. Its exact size and column details are not specified in the provided metadata.

Use Cases

• Training classifiers to detect hate speech based on the three-class labeling scheme.
• Benchmarking model performance on the distinction between offensive language and hate speech.
• Studying linguistic patterns and prevalence of toxic content on social media platforms.
• Developing multi-class text classifiers for content safety applications.

Strengths

• Dataset is based on a peer-reviewed paper published at ICWSM (AAAI) in 2017.
• Provides a clear three-class taxonomy (hate speech, offensive, neither) for model training.
• Original data and paper URLs are provided for verification and further study.

Limitations

• Row count, column details, and file formats are unknown, limiting suitability assessment.
• Data is from 2017 and may not reflect current language patterns on social media.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Department of Sociology, Cornell University, Ithaca, NY, USA

Collection Method

Tweets were collected and annotated, as described in the paper 'Automated Hate Speech Detection and the Problem of Offensive Language'.

Time Range

null

Freshness

Last update date is unknown; freshness unverified.

Geography

null