Extract readable PDFs style information

[grossir]

We already use pdfplumber to extract a readable PDF's text. If we use page.extract_words in conjunction with page.extract_text it will sometimes be able to extract style information. This may improve readability and help on issues like freelawproject/eyecite#198 where style tags <i> and <em> are important.

doctor/doctor/lib/text_extraction.py

Lines 67 to 69 in 9e0e76f

	page_text = page.extract_text(
	layout=True, keep_blank_chars=True, y_tolerance=5, y_density=25
	)

For example, to get the italics in the second page

import requests
import pdfplumber
from io import BytesIO

url = "https://storage.courtlistener.com/pdf/2025/01/30/georgia_insurers_insolvency_pool_v._logisticare_solutions_llc.pdf"
r = requests.get(url)
pdf = pdfplumber.open(BytesIO(r.content))
words = pdf.pages[1].extract_words(extra_attrs=["fontname"])

In [33]: [i['text'] for i in pdf.pages[1].extract_words(extra_attrs=["fontname"]) if i['fontname'] == 'TKMUCK+EquityARegular,Italic']
Out[33]: ['Wade', 'v.', 'Allstate', 'Fire', '&', 'Cas.', 'Co.']

Then, the italicized words can be resolved to the extracted main text. We would probably need to prebuild a list of courts where this is possible, and filter what styles we want to be preserved.

[grossir]

Another example where extracting styles (and linking the reference citations) would greatly improve readability