5,000,000 questions and 3,000,000 answers collected from Google search autocomplete and answer boxes. The dataset features naturalistic, short-form queries paired with diverse response formats including short text, long text, and structured collections.
Use Cases
- Train question-answering systems using the question and answer pairs to mimic search engine behavior
- Analyze query patterns using the 5 million naturalistic questions collected from autocomplete
- Classify response structures by distinguishing between textual answers and structured collections
Strengths
- 5,000,000+ questions generated semi-automatically from Google autocomplete
- 3,000,000+ answers extracted from Google search answer boxes
- Diverse answer types including short text, long text, and structured collections