Quora Question Triplets for Semantic Equivalence Training

Triplets of Quora questions for training semantic equivalence models. It is structured with anchor, positive, and negative question examples. The dataset size is categorized as between 100,000 and 1,000,000 rows.

Use Cases

• Train a Sentence Transformer model using anchor, positive, and negative question triplets for semantic similarity tasks.
• Fine-tune a model for semantic similarity classification on English text using the triplet structure.
• Benchmark contrastive learning algorithms on question pairs derived from Quora data.
• Analyze patterns in semantically equivalent (positive) and non-equivalent (negative) question formulations.

Strengths

• Dataset size is categorized as between 100,000 and 1,000,000 rows, providing substantial data for training.
• Data is specifically structured as triplets (anchor, positive, negative) for contrastive learning tasks.
• The dataset is tagged for the semantic similarity classification task and English language modality.

Limitations

• The exact number of rows, columns, and sample data are unavailable, limiting precise assessment of scale and structure.
• The original Quora team did not author the dataset card, and provenance details are incomplete.
• Data is region-tagged as US, which may limit generalizability to other English dialects or languages.

Provenance

Source

Quora, curated and uploaded by the Hugging Face team under the account 'embedding-data'.

Collection Method

Organized as triplets (anchor, positive, negative) from Quora data; specific gathering method is unknown.

Freshness

Last updated on 2022-08-02.

Geography

Region is tagged as US.