google-research-datasets/natural-questionsArchived
Natural Questions
Natural Questions is a large-scale machine learning research dataset designed for training and evaluating open-domain question answering systems. It consists of a corpus of real search queries paired with human-annotated Wikipedia document spans, providing a standardized foundation for advancing automated information retrieval and comprehension technologies.
The project distinguishes itself by providing high-quality ground truth data that supports multiple answer formats, including binary, short-form, and long-form responses. By incorporating extractive span annotations and structured document representations, it enables models to learn how to locate precise information within complex source materials rather than relying on simple text matching.
The dataset includes comprehensive tooling for benchmarking and validation, allowing researchers to measure model performance against verified reference data. It supports the entire lifecycle of question answering development, from training models on annotated queries to executing automated evaluation pipelines that ensure consistent and reproducible results.
Features
- Question Answering Datasets - Serves as a comprehensive dataset of search queries and Wikipedia document spans for training and benchmarking question answering systems.
- Question Answering Benchmarks - Serves as a standardized benchmark dataset for evaluating open-domain question answering models.
- Machine Learning Datasets - Provides a large-scale corpus of real search queries and annotated document spans for training and evaluating machine learning models.
- Question Answering Models - Provides the foundational data and tooling required to train and refine question answering systems.
- Precomputed Ground Truth Datasets - Provides high-quality, human-annotated ground truth datasets for training and benchmarking models.
- Evaluation Pipelines - Provides automated evaluation pipelines that score model inference outputs against verified reference data.
- Document Information Extraction - Identifies and marks precise text segments within documents to locate information.
- Natural Language Processing Datasets - Contains a large-scale collection of annotated text spans and queries designed to improve automated comprehension and information retrieval.
- Model Performance Metrics - Calculates performance metrics to validate model output quality against reference fields.
- Model Predictions - Generates structured model predictions from input data for performance benchmarking.
- Natural Language Processing - Advances information retrieval technologies through the analysis of real search queries and document responses.
- Answer Span Prediction - Predicts the start and end positions of answers within source documents.
- Span Extraction Utilities - Identifies and extracts specific text spans from documents to serve as precise answers.
- Model-to-Image Packaging - Bundles models and dependencies into container images for consistent evaluation.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
