Make Document Intelligence ignore redundant info

Question

Hiya

I'm currently working a program in C# that tries to index documents and put them in a search service index.
These documents are pdfs and word files, with text and screenshots explaining the text. All these documents go through a Document Analysis instance for interpreting the content of these files.

I was wondering, is there a possibility to ignore only the screenshots, since the instance picks up words out of these screenshots. These words are pretty much completely redundant.

Thanks in advance!

Answer 1

@Thibault Verlinde - Thanks for reaching out to us.

It is possible to ignore screenshots when analyzing documents using a Document Analysis instance. One way is to use an object detection model to identify and classify images within the document. Once the images are identified, they can be excluded from the analysis.

Another approach is to use a layout analysis model to extract text, tables, and other elements while ignoring images. Azure’s Document Intelligence (formerly Form Recognizer) offers layout analysis capabilities that can be used for this purpose

Here is the document for more reference: https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/concept-layout?view=doc-intel-4.0.0

https://learn.microsoft.com/en-us/python/api/azure-ai-formrecognizer/azure.ai.formrecognizer.documentanalysisclient?view=azure-python

Hope this helps. and please feel free to reach out if you have any further questions.

---

If the above response was helpful, please feel free to "Accept as Answer" and click "Yes" so it can be beneficial to the community.