# google-research-datasets/natural-questions

Natural Questions (NQ) contains real user questions issued to Google search, and answers found from Wikipedia by annotators. NQ is designed for the training and evaluation of automatic question answering systems.

Repository: https://github.com/google-research-datasets/natural-questions
Canonical: https://ross.abutalabs.com/products/natural-questions
Language: Python
License: Apache-2.0
License Family: permissive
Archived: true
Last push: 2021-07-30T00:22:08+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2780, "days_push": 1861, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1138, forks 164 (observed 2026-08-28T04:03:44.083816+00:00)

## What it is
Natural Questions is a large-scale question answering dataset from Google containing 307k real user search questions paired with Wikipedia pages annotated with long and short answers. The repository provides data utilities, a data browser, and evaluation tooling for training and benchmarking QA systems.

## Use cases
- train a question answering model on real search queries
- evaluate extractive QA systems against a benchmark
- get question-answer pairs from Wikipedia pages
- benchmark long answer and short answer selection models
- compare my QA system on the Natural Questions leaderboard

## When to choose
- you need a large, realistic QA benchmark based on real user queries
- you want to compare against published QA baselines and leaderboard results
- you need Wikipedia-sourced training data with human-annotated answers

## When to avoid
- you need conversational or multi-turn QA data
- you need small lightweight datasets for quick prototyping
- you need non-English question answering data

## Facets
- artifact type: dataset
- maturity: maintenance
- function: machine-learning, nlp, data-science
- domain: machine-learning
- platform: python
- tags: question-answering, benchmark, wikipedia, google-search-queries, dataset, natural-language-processing, search

## Member repositories
- google-research-datasets/natural-questions (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:44.083816+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:36:09.031073+00:00, confidence not recorded.
  - readme: https://github.com/google-research-datasets/natural-questions (fetched 2026-08-28T04:03:44.083816+00:00, sha 9ca80dc605ab)
  - registry_pypi: https://pypi.org/pypi/natural-questions/json (fetched 2026-08-29T12:41:16.727988+00:00, sha 26b41aa48518)
- Data as of 2026-08-30T08:39:29.467469+00:00.