Document Visual Question Answering Examples with Images and Text

DocVQA 1200 Examples is a multimodal dataset for visual question answering on documents. It contains 1,200 examples of images paired with text and questions, created by author nielsr and last updated in August 2022.

Use Cases

• Train a model to answer questions about document content using the provided image and text data.
• Benchmark the performance of multimodal models on the specific task of Document Visual Question Answering (DocVQA).
• Fine-tune a vision-language model using the image-question-answer triplets to improve document comprehension.
• Analyze the relationship between visual document elements and textual questions to understand model reasoning.

Strengths

• Contains 1,200 curated examples for a specific task.
• Integrates both image and text modalities for each example.

Limitations

• Relatively small size of 1,200 examples may limit model training potential.
• Dataset is several years old, potentially not reflecting latest document formats or tasks.

Provenance

Source

huggingface

Collection Method

Curated collection for Document Visual Question Answering.

Freshness

Last updated in August 2022.

Geography

Region tags suggest a US focus, but specifics are unknown.