# Docmatix - a huge dataset for Document Visual Question Answering

DevFeed: [Docmatix - a huge dataset for Document Visual Question Answering](<https://devfeed.tech/articles/docmatix-a-huge-dataset-for-document-visual-question-answering-7173.md>)

Original publisher: [Read original article](<https://huggingface.co/blog/docmatix>)

Author: Andres Marafioti; Hugo Laurençon

Published: 2024-07-18T00:00:00Z

Content type: article

Language: en

Sources: [Hugging Face - Blog](<https://devfeed.tech/sources/hugging-face-blog.md>)

Topics: [datasets](<https://devfeed.tech/topics/datasets.md>), [vlm](<https://devfeed.tech/topics/vlm.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [hugging face](<https://devfeed.tech/topics/hugging-face.md>), [pdf](<https://devfeed.tech/topics/pdf.md>), [Language models](<https://devfeed.tech/topics/language-models.md>)

Tags: [announcement](<https://devfeed.tech/tags/announcement.md>), [code](<https://devfeed.tech/tags/code.md>), [community](<https://devfeed.tech/tags/community.md>), [cv](<https://devfeed.tech/tags/cv.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [hallucinations](<https://devfeed.tech/tags/hallucinations.md>), [hugging-face](<https://devfeed.tech/tags/hugging-face.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [pdf](<https://devfeed.tech/tags/pdf.md>), [research](<https://devfeed.tech/tags/research.md>), [synthetic-data](<https://devfeed.tech/tags/synthetic-data.md>), [vlm](<https://devfeed.tech/tags/vlm.md>)

## AI overview

The article introduces Docmatix, a large Document Visual Question Answering dataset containing 2.4 million images and 9.5 million question-answer pairs derived from 1.3 million PDF documents. It describes the dataset's generation from PDFA transcriptions using a Phi-3-small model, quality filtering, image processing, and prompt optimization.

## Source excerpt

We're on a journey to advance and democratize artificial intelligence through open source and open science.