Search PDFs with Transformers and Python Notebook

Re:Search PDFs with Transformers and Python Notebook by shubham_saboo on Monday July 25, 2022 @10:17AM (#32224615)

Wao, this is a really cool way to build full fledged search that too in a notebook!

Does it work end-to-end with PDF as a data structure or do we have to use OCR and parse the text first to be able to search it, really curious?

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Monday July 25, 2022 @10:20AM (#32224654)

The version in the notebook is just for simple text-based PDFs. I wrote some posts on our company blog[1] about the sheer agonies of dealing with PDF as a data format, so wanted to stick with as simple as possible for now.

That said, I'm planning future notebooks where you can perform text-to-image or image-to-image search, integrate OCR, scale it up, serve it, deploy it, etc.

[1] https://medium.com/jina-ai

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by shubham_saboo on Monday July 25, 2022 @10:21AM (#32224678)

Awesome, will be on the lookout for that!

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Monday July 25, 2022 @10:23AM (#32224698)

We've got quite a few other notebooks for other kinds of search on the blog. Would love to hear your thoughts!

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by rahimnathwani on Monday July 25, 2022 @10:22AM (#32224693)

Under the hood, it uses https://github.com/pdfminer/pdfminer.six which expects the text to be stored as text.

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Monday July 25, 2022 @10:28AM (#32224779)

You mean the PDFSegmenter Executor in the notebook?

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by rahimnathwani on Monday July 25, 2022 @10:33AM (#32224852)

Yes

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Monday July 25, 2022 @11:12AM (#32225494)

PDFSegmenter also extracts images, which can then be OCR'ed in the next step of the pipeline

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by spaetzleesser on Monday July 25, 2022 @04:55PM (#32230798)

"PDF as a data structure"

Don't. PDF is a terrible format for storing machine readable data. You lose a ton of Information while you create the PDF which you then painstakingly have to get back later (if that's even possible)

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Monday July 25, 2022 @05:57PM (#32231473)

I may have misworded it (if I wrote those words - PDF rots the brain and my memory likewise).

Agreed on the rest. PDFs don't store machine-readable data. Often just pixelated scanned hot garbage dumpster fire text.

I hate PDFs but have to work with the satanforesaken things. Hence the notebook. It's my little way of trying to give my little PDF-bespoked-hellscape a tiny little glow-up.

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by spaetzleesser on Monday July 25, 2022 @10:40PM (#32233639)

I probably didn’t read your comment closely enough. When I hear about PDF parsing or PDF as data I immediately get flashbacks from a project years ago where I had to parse PDF files. I think I am still traumatized by this experience so whenever I hear somebody wants to do this I just want to scream “Nooo. Don’t do this”

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Tuesday July 26, 2022 @04:14AM (#32235656)

I think you and I should start a support group!

[ Parent | on HN ]

Re:Search PDFs with Transformers and Python Notebook by alexcg1 on Monday July 25, 2022 @11:43AM (#32225991)

Incidentally Jina Hub [0] has a few OCR Executors [1][2] you could integrate into my notebook (though you'd have to do some rewiring to take images into account since it's a text-based notebook)

[0] https://hub.jina.ai/

[1] https://hub.jina.ai/executor/w4p7905v

[2] https://hub.jina.ai/executor/78yp7etm

[ Parent | on HN ]