Natural Questions Slim Short Answer: Flattened Question-Answer Pairs for QA Experiments

A slim, flattened derived version of the Natural Questions dataset for short-answer question answering experiments. The dataset, created by BOB12311, removes original document HTML and long answer candidates to provide simple question-answer pairs. It was last updated on the platform on 2026-05-17.

Use Cases

• Fine-tuning language models for short-answer extraction based on the simplified question-answer structure.
• Benchmarking QA system performance on a flattened, answer-focused dataset.
• Prompt engineering experiments using clean question-answer pairs without document context.

Strengths

• Derived from the established 'google-research-datasets/natural_questions' source.
• Simplified structure intended for lightweight QA prompting and evaluation.
• Last updated on 2026-05-17, indicating recent maintenance.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Derived from google-research-datasets/natural_questions.

Collection Method

Conversion that kept examples with extractable short answers and removed original document HTML and long answer candidates.

Time Range

null

Freshness

Last updated 2026-05-17 13:24:41.

Geography

null