Going paperless. It’s a conundrum as it usually creates just as much work as it tries to solve. Perhaps applying machine learning to the documents will help ease the workflow.

For the last day, on and off, I’ve been training classifiers1 for Paperless-ng, an open source software to help manage scans of physical documents. It took a little bit of brain-wrapping but I managed to work out a workflow.

Previously I had scanned all my receipts and what-not in batches and manually renamed the files by date and correspondent. Now I’ll just dump the scans straight into Paperless and let the AI do its thing. Then I’ll fix as necessary. Also there’s full text search, so that’s useful.

I want to get a scanner that can sit on the counter in the kitchen that doesn’t need to be connected to a computer. It will scan documents directly to the server, and the server will try to figure out the who, what, and when of the document. This should be a one-button thing where you don’t have to think about it and, as long as the scan goes through, you can access it from your phone or computer whenever or wherever you need to. This way any family member can scan documents into the system. Insert paper, press button, walk away.

Legal documents, receipts, instruction manuals, and all the paper crap that we get bombarded with every day. It’s nice to have computerized and categorized access to the information without the hassle of storing the physical medium.


  1. “Classifiers” is computer-speak for how the machine learning (i.e., AI) algorithms model their behavior. I think. ↩︎