Skip to content

Dataset · google

BoolQ: yes/no text-based questions

google/boolq

15,942 natural yes/no questions based on an English text snippet.

What's inside

BoolQ is an English-language dataset of questions that can be answered only with yes or no. Each example is a triple (question, passage, answer); the page title may be included as additional context. The questions were collected under natural conditions without prompts, totaling 15,942 examples. The task is close to text pair classification and NLI.

What you can do with it

  • Training and evaluating question-passage classifiers
  • Testing yes/no QA models
  • A benchmark for NLI-like models
Tasks
Text classification
Languages
English
Size
10K–100K rows
Downloads this month
68.4K
Likes
124

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("google/boolq")