How can I enable users to label PDF documents directly in my application without using Document Intelligence Studio?

Question

I want to implement the labeling process within my application, so users don’t need to visit Document Intelligence Studio. My goal is to generate the required <filename>.pdf.labels.json and fields.json files directly in my app. How can I achieve this?

Answer 1

You can build a custom interface that allows users to upload and view PDFs, select text or regions, and assign labels. Use tools like PDF.js to render PDFs and capture user annotations, such as bounding boxes or text selections. Save these annotations as JSON files (<filename>.pdf.labels.json and fields.json) that match the format required by Azure AI Document Intelligence. This way, users can label PDFs without leaving your app, and you can use these files for training or inference directly.