Parallelize Azure Document Intelligence API call

Mohamed Yasser 0

I am wondering if i can speed up my document intelligence function, as i searched alot but in vain. I tried to cut my pdf to chunks and parallelize the running of the function but it didn't go faster. Do Azure provide a way of multi-processing or parallelization of the Azure AI Document Intelligence.

SriLakshmi C 2,845 Reputation points Microsoft Vendor

2025-02-25T11:56:02.74+00:00
Hello Mohamed Yasser,

Greetings and Welcome to Microsoft Q&A!

Azure Functions can help with multi-parallel processing of Document Intelligence PDFs. You can use Azure Functions to create a serverless architecture that can scale automatically to handle large volumes of PDFs.

Here's how you can use Azure Functions to process PDFs in parallel:

Split the PDF into chunks of 30 pages each, as you have already done.

Upload the PDF chunks to Azure Blob Storage.

Create an Azure Function that listens to the Blob Storage container for new PDF chunks.

When a new PDF chunk is uploaded, trigger the Azure Function to process the chunk using the Document Intelligence API.

Use Azure Functions' ability to scale automatically to process multiple PDF chunks in parallel.

Once all the PDF chunks have been processed, reassemble them into the original PDF.

I hope this helps. Thank you.
SriLakshmi C 2,845 Reputation points Microsoft Vendor

2025-02-26T08:58:46.9566667+00:00

Hi Mohamed Yasser,

Did you get any chance to check the above response. Thank you!
Mohamed Yasser 0 Reputation points

2025-02-26T10:28:25.2133333+00:00

Pretty good idea, but I have a concern. What if the PDF have tables inside, and while splitting the pdf, a table was also splitted the first half is in the last pdf split and the second half is in the first page of the other pdf split, aren't by that way i will lose semantic meaning of the table analyzation?
SriLakshmi C 2,845 Reputation points Microsoft Vendor

2025-02-28T07:40:48.9866667+00:00

Hi Mohamed Yasser,

Yes, splitting PDFs that contain tables can lead to the loss of semantic meaning in structured data, as table rows might get separated between different chunks. Azure Document Intelligence does not have built-in support for preserving table structures across multiple PDF chunks. To address this,

Consider using Smart PDF Splitting with Table Detection to ensure tables remain intact during document segmentation.

Also, you can process the entire PDF and extract tables separately to maintain data consistency. If splitting is unavoidable, implement post-processing techniques to merge split tables, ensuring the extracted data retains its original structure and meaning.

Thank you!

Share via

Parallelize Azure Document Intelligence API call

Your answer