I am trying to extract the information from the following but with no luck so far. Here is the link to the test PDF that I am using I'm hoping that if I can come up with a strategy for this single page I can extrapolate that over a large collection set. import pdfplumber import pandas as pd pdf_path = "/Users/Blacky_1_2/Desktop/python/nominal_rolls/pdf/e011087784.pdf" output_csv = "/Users/Blacky_1_2/Desktop/python/nominal_rolls/csvs/2nd.csv" # Open the PDF with pdfplumber.open(pdf_path) as pdf: p1 = pdf.pages[0] settings = { "vertical_strategy": "text", "horizontal_strategy": "text", "snap_y_tolerance": 1, "intersection_x_tolerance": 3, } table = p1.find_table(table_settings=settings) im = p1.to_image(resolution=150) im.reset().debug_tablefinder(settings) #table = pd.DataFrame(page_crop_new.extract_table(settings)) display(im) # show the debug image I've tried messing around with various settings, but no matter which combination I choose the finding the table doesn't seem to be identifying the cutoff of one field with the next on a handful of the columns. If I use the vertical and horizontal_strategy as both "text" this is what the result looks like in the Jupyter Notebook: When I adjust the horizontal_strategy = "lines" in the settings it is easier to see where the separation of one column is compared to the next (which is obviously incorrect): I have tried to see if I could explicitly state the vertical lines as well as identify and disregard the headers, but no luck. I've played around with other modules (import fitz # PyMuPDF) but the regex seems to be having difficulty outside of the basic content and is also not finding the row contents cleanly. I realize the PDF is of an older scan and may be difficult to parse, but I feel like the quality is good enough to be able to extract the appropriate information. Continue reading...
