3 ms·

okay but what if my PDF is non parseable? Not sure if Python's any good for that

alexcg1 4y ago
In that case I'd use: 1. PDFSegmenter (in the notebook) - extract the images of the text (yup, it does images too) 2. An OCR Executor [0][1] from Jina Hub [2] to extract the text from the images 3. Actually splice the text chunks together to be what you'd expect - that's the tricky part. Even text splitting over pages can be tricky to reassemble properly. PDFs are a pain the butt frankly. [0] https://hub.jina.ai/executor/78yp7etm https://hub.jina.ai/executor/78yp7etm [1] https://hub.jina.ai/executor/w4p7905v https://hub.jina.ai/executor/w4p7905v [2] https://hub.jina.ai https://hub.jina.ai
nicodjimenez 4y ago
Mathpix PDF search is fully visually powered and does not use underlying PDF metadata, even working on handwriting. It’s a great choice for researchers (especially in STEM) who want to build a searchable archive of PDFs.
simonw 4y ago
Amazon Textract does a phenomenal job of extracting text from dodgy scanned PDFs - I've been running it against scanned typewritten text and even handwritten journal text from the 1880s with great results. I built a tool for running OCR against every PDF in an S3 bucket (which costs about $1.50/thousand pages) here: https://simonwillison.net/2022/Jun/30/s3-ocr/ https://simonwillison.net/2022/Jun/30/s3-ocr/
- ydant 4y ago
  Textract really does do a good job of balancing cost, ease of use, and quality, at least for my hobbyist needs. I was inspired by another recent comment you posted on HN, and after some testing of the Textract console [0] I wrote a simple "local only" command-line version [1] (Python, boto3) that does similar things to your tool. I used my tool to OCR a few hundred comic strip images I've been meaning to OCR for a while now - the service did beautifully where other tools I've tried in the past struggled with the handwritten text on the comics. Textract is fast enough that running serially was fine for a one-off without involving the more parallelized S3 workflow. [0] https://us-east-1.console.aws.amazon.com/textract/home?region=us-east-1#/demo https://us-east-1.console.aws.amazon.com/textract/home?regio... [1] https://github.com/mbafford/textract-cli https://github.com/mbafford/textract-cli
  - simonw 4y ago
    That's really neat: https://github.com/mbafford/textract-cli/blob/master/textract_cli/textract_cli.py https://github.com/mbafford/textract-cli/blob/master/textrac... - my tool had to involve an S3 bucket too just because Textract won't let you upload a PDF directly to it without storing it in S3 first.
    - ydant 4y ago
      Based on some testing just now, it looks like for synchronous mode single-page PDF is supported, but if you try to OCR a multi-page document, it throws: An error occurred (UnsupportedDocumentException) when calling the DetectDocumentText operation: Request has unsupported document format Normally, I just use pdfsandwich [0] for PDFs, which has been good enough generally, but I'm tempted to switch to using textract, because it's been much better at rough scans for my tests. [0] http://www.tobias-elze.de/pdfsandwich/ http://www.tobias-elze.de/pdfsandwich/

In that case I'd use: 1. PDFSegmenter (in the notebook) - extract the images of the text (yup, it does images too) 2. An OCR Executor [0][1] from Jina Hub [2] to extract the text from the images 3. Actually splice the text chunks together to be what you'd expect - that's the tricky part. Even text splitting over pages can be tricky to reassemble properly. PDFs are a pain the butt frankly. [0] https://hub.jina.ai/executor/78yp7etm https://hub.jina.ai/executor/78yp7etm [1] https://hub.jina.ai/executor/w4p7905v https://hub.jina.ai/executor/w4p7905v [2] https://hub.jina.ai https://hub.jina.ai

nicodjimenez 4y ago

Mathpix PDF search is fully visually powered and does not use underlying PDF metadata, even working on handwriting. It’s a great choice for researchers (especially in STEM) who want to build a searchable archive of PDFs.

simonw 4y ago

Amazon Textract does a phenomenal job of extracting text from dodgy scanned PDFs - I've been running it against scanned typewritten text and even handwritten journal text from the 1880s with great results. I built a tool for running OCR against every PDF in an S3 bucket (which costs about $1.50/thousand pages) here: https://simonwillison.net/2022/Jun/30/s3-ocr/ https://simonwillison.net/2022/Jun/30/s3-ocr/

ydant 4y ago
Textract really does do a good job of balancing cost, ease of use, and quality, at least for my hobbyist needs. I was inspired by another recent comment you posted on HN, and after some testing of the Textract console [0] I wrote a simple "local only" command-line version [1] (Python, boto3) that does similar things to your tool. I used my tool to OCR a few hundred comic strip images I've been meaning to OCR for a while now - the service did beautifully where other tools I've tried in the past struggled with the handwritten text on the comics. Textract is fast enough that running serially was fine for a one-off without involving the more parallelized S3 workflow. [0] https://us-east-1.console.aws.amazon.com/textract/home?region=us-east-1#/demo https://us-east-1.console.aws.amazon.com/textract/home?regio... [1] https://github.com/mbafford/textract-cli https://github.com/mbafford/textract-cli
- simonw 4y ago
  That's really neat: https://github.com/mbafford/textract-cli/blob/master/textract_cli/textract_cli.py https://github.com/mbafford/textract-cli/blob/master/textrac... - my tool had to involve an S3 bucket too just because Textract won't let you upload a PDF directly to it without storing it in S3 first.
  - ydant 4y ago
    Based on some testing just now, it looks like for synchronous mode single-page PDF is supported, but if you try to OCR a multi-page document, it throws: An error occurred (UnsupportedDocumentException) when calling the DetectDocumentText operation: Request has unsupported document format Normally, I just use pdfsandwich [0] for PDFs, which has been good enough generally, but I'm tempted to switch to using textract, because it's been much better at rough scans for my tests. [0] http://www.tobias-elze.de/pdfsandwich/ http://www.tobias-elze.de/pdfsandwich/