I'm looking to develop a web application that can parse paper documents like invoices and purchase orders, which might include handwritten notes. I'm curious about how well AWS Textract performs when it comes to reading handwriting in real-world scenarios. If a human can struggle with messy handwriting, how does Textract's accuracy compare when interpreting such text?
4 Answers
I’ve heard that the handwriting recognition tech in devices like Kindle Scribe is quite good, so I assume Textract uses similar algorithms. It’s impressive, but I’d love to know if it makes you confirm deletions by typing the word 'delete' too when you’re processing!
I've found Textract to be pretty solid overall. Why not give it a shot with some of the documents you plan on working with? You might be surprised by the results!
I’ve used Textract and found it reasonably accurate, especially when it comes to standard printed texts. They provide confidence scores, making it easy to spot potential errors, which you could then double-check with a human if needed.
In my experience, Textract offers decent accuracy with printed text, but I found it less reliable with handwritten notes. The costs are quite high too; it can be around $50 for every 1000 invoices processed. Honestly, I think using models like openrouter with GPT-5 or alternatives could provide better results at a lower price.
How do those models perform with really messy handwriting, though? Just curious if they're worth it!
Haha, that would be a bit excessive, but you’d know they’re Amazon developers for sure!