topic Re: Auto-extract text from attached documents in ArcGIS Pro Questions

Auto-extract text from attached documents

Shane_EU — Tue, 09 Jul 2024 06:12:15 GMT

Hi, does anybody have any ideas how could this be implemented in ArcGIS Pro 3.3? Thanks!

Re: Auto-extract text from attached documents

jcarlson — Tue, 09 Jul 2024 12:47:23 GMT

I think you'll want to look at the textract module: https://textract.readthedocs.io/en/stable/

I haven't used it in quite the way you've specified, but I imagine it could be fairly straightforward with some Python.

I'm more comfortable with the ArcGIS Python API than ArcPy. Is this process meant to edit local file-based resources? AGOL Layers? It sounds like you're editing a geodatabase directly, given the reference to the __ATTACH table.

The ArcGIS Python API has a helpful submodule for working with attachments, and you could probably do it in a couple of for loops.

Re: Auto-extract text from attached documents

Shane_EU — Tue, 09 Jul 2024 13:03:16 GMT

Thanks Josh. Yes it's for a local file-based GDB not AGOL or Enterprise, and should happen at the point of attaching the document. I'll look into textract 👍

Re: Auto-extract text from attached documents

AdamMartin — Fri, 14 Feb 2025 14:02:53 GMT

@Shane_EU I'm sure there's a way to do this in ModelBuilder but I do not have an example I can share. @BruceHarold may have an example using the Data Interop Extension for Pro. I don't know the specific interactions these tools have with feature Attachments, but have you also explored extracting text objects as entities (rather than the full text) into properties? Have you explored the Text Classifications tools using AI (https://pro.arcgis.com/en/pro-app/3.3/tool-reference/geoai/how-entity-recognition-works.htm) and LocateXT which can extract locations, but also text strings using regex patterns? https://pro.arcgis.com/en/pro-app/3.3/help/data/locatext/extract-locations.htm @JeffWilson4

Re: Auto-extract text from attached documents

BruceHarold — Fri, 14 Feb 2025 14:12:19 GMT

I think this could be modeled or scripted, but it might make more sense to use the LocateXT tools on the incoming Word documents before attaching them, to make accessory attachments that go for the ride. Data Interoperability doesn't add anything in this situation, as described, but might if other complexities arise.

Re: Auto-extract text from attached documents

JeffWilson7 — Fri, 14 Feb 2025 17:19:18 GMT

@Shane_EU @AdamMartin @BruceHarold This would be super easy to do with LocateXT. So you have word docs attached to features? How are the docs related to the features? What are the features? Do the documents identify a location in the text that is the feature? This would be super easy to do either way, but there might be a more impressive workflow if understood the basics of what you are trying to do.