Tweet Paraphrase Pairs from SemEval 2015 Workshop

Paraphrase pairs of tweets created for the SemEval 2015 workshop, intended for evaluating text embedding models. It is part of the Massive Text Embedding Benchmark (MTEB) and is categorized for text-to-text tasks within social and written domains.

Use Cases

• Evaluate embedding models on the semantic similarity classification task using tweet paraphrase pairs.
• Benchmark paraphrase detection performance on social media text from the SemEval 2015 workshop.
• Train models for text classification tasks using human-annotated tweet pairs.

Strengths

• Human-annotated data from a defined academic workshop (SemEval 2015).
• Part of the standardized Massive Text Embedding Benchmark (MTEB) for model evaluation.
• Monolingual English text focused on social media domain.

Limitations

• Specific row count, column details, and sample data are unavailable for assessment.
• Dataset size is categorized as 'n1 K', indicating a relatively small scale.
• Temporal coverage is limited to the context of the 2015 workshop, potentially lacking recent data.

Provenance

Source

SemEval 2015 workshop (Task 1).

Collection Method

Human-annotated paraphrase pairs of tweets.

Time Range

Context from 2015 workshop.

Freshness

Last metadata update was 2025-05-04, but the underlying data originates from 2015.

Geography

Region tag indicates 'us', but specific coverage is unknown.