identify rectangle content (its not a word, image, curve)

[pauljohn32]

I'm practicing with pdfplumber to detect fields in US government forms. Example today is a tax form. These have lots of checkboxes.

Typically, the pdf are generated by tax prep software. The checkboxes are not Acrobat form elements, so far in my experience. I attach an example IRS form for testing.

f1040.pdf

.extract_text and .extract_words work as expected for me, except for these checkbox things. These are not extracted as text or words. These can be found in the rects for the page. So I know there's something in there, but what is it?

The visualization tool works, using code below I show highlighted check boxes with ImageMagick tool in pdfplumber.

With the attached f1040.pdf file, I run these to make the picture:

irs =  pdfplumber.open(fn1)
irs1 = irs.pages[0]
irs1rects = irs1.rects
im = irs1.to_image(resolution=150)
irs1checkrects = [j for j in irs1rects if (j['bottom'] > 75) and (j['bottom'] < 77)]
im.draw_rects(irs1checkrects)

If I draw the extract_words output, the checkboxes are not selected. That's why I dig through the rects to find the coordinates that align. I hoped that the checkbox itself and the X mark inside it (if there is one) will be some Unicode characters.

I appreciate your advice.

[jsvine]

Hi @pauljohn32, and thanks for sharing this interesting example. Although these checkboxes are not Acrobat form elements, they are another kind of interactive interactive element, an "annotation." These can be accessed via page.annots.

In the linked and attached (discussion-641.ipynb.txt) notebook, you can see what the data for those annotations look like, and how they change when you've checked the checkboxes (as I have in this file, also referenced in the notebook: f1040-checked.pdf).