Dataset · google
Question answering on charts and tables from scientific papers
google/spiqa
Google's SPIQA dataset for evaluating question-answering systems on scientific paper illustrations, charts, tables, and text.
What's inside
The dataset contains questions and answers based on charts, tables, and text snippets from computer science research papers. Although described as a large benchmark, the current export metadata indicates a size of under 1,000 records and a text modality. The project was published by the Google team along with evaluation metric code.
What you can do with it
- Testing multimodal models on understanding scientific charts and diagrams
- Evaluating information extraction systems on complex technical documents
- Benchmarking QA models in the computer science domain
- Tasks
- Question answering
- Size
- up to 1K rows
- Downloads this month
- 3.4K
- Likes
- 51
How to load it
from datasets import load_dataset
ds = load_dataset("google/spiqa")