PDF 다운로드

!wget https://www.mhlw.go.jp/content/10601000/000853766.pdf -O data.pdf

프로그램

import pdfplumber
import pandas as pd

from tqdm.notebook import tqdm

with pdfplumber.open("data.pdf") as pdf:

    dfs = []

    for page in tqdm(pdf.pages[1:]):

        # 縦線
        vertical = [
            edge["x0"]
            for edge in page.debug_tablefinder().edges
            if edge["orientation"] == "v"
        ]

        # 縦線の一番下の位置
        bottom = max(
            [
                edge["bottom"]
                for edge in page.debug_tablefinder().edges
                if edge["orientation"] == "v"
            ]
        )

        # 横線
        horizontal = [
            edge["top"]
            for edge in page.debug_tablefinder().edges
            if edge["orientation"] == "h"
        ]

        # 横線に縦線の一番下の位置を追加
        horizontal.append(bottom)

        table_settings = {
            "vertical_strategy": "explicit",
            "explicit_vertical_lines": vertical,
            "horizontal_strategy": "explicit",
            "explicit_horizontal_lines": horizontal,
            "snap_tolerance": 3,
            "intersection_tolerance": 3,
        }

        table = page.extract_table(table_settings)

        df_tmp = pd.DataFrame(table)

        dfs.append(df_tmp)

df = pd.concat(dfs)

df.shape

df.to_csv("result.csv", encoding="utf_8_sig")