df16c98366
공공기관2 작업 중. _temp 몽타주(재생성가능)는 제외. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
113 lines
4.3 KiB
Python
113 lines
4.3 KiB
Python
# -*- coding: utf-8 -*-
|
|
"""공공기관 N 영상/오디오 거짓양성 교정.
|
|
원인: phase234 detect_media가 str(body)에 '.mp4'/'.mp3'만 있어도 영상/오디오로 오판(템플릿 배경영상/플레이어 스크립트).
|
|
교정: 영상=youtube/vimeo iframe·링크 또는 <video>실소스 / 오디오=<audio>실소스·미디어링크 만 인정. 게시판은 상세 표본도 확인.
|
|
대상: N에 '영상' 또는 '오디오' 포함 & L!=사이트 행. 거짓이면 N에서 제거.
|
|
사용: python _공공기관_navfix.py [기관명 ...]
|
|
"""
|
|
import sys, os, re, json, importlib.util, warnings
|
|
import openpyxl
|
|
warnings.filterwarnings('ignore')
|
|
spec = importlib.util.spec_from_file_location('p234', r'D:\01.프로젝트\DB수집\_스크립트\_공공기관_phase234.py')
|
|
P = importlib.util.module_from_spec(spec); spec.loader.exec_module(P)
|
|
OUTDIR = r'D:\01.프로젝트\DB수집\공공기관'
|
|
# 실제 콘텐츠 영상만: watch/embed/youtu.be ID/vimeo 숫자. ⛔채널·계정·SNS링크(/channel//user//@//c/) 제외
|
|
VID = re.compile(r'(youtube\.com/watch\?|youtube\.com/embed/|youtu\.be/[\w-]{6,}|player\.vimeo\.com/video|vimeo\.com/\d{5,})', re.I)
|
|
AUD_EXT = re.compile(r'\.(mp3|wav|m4a|ogg|flac)(\?|$|["\'&])', re.I)
|
|
|
|
|
|
def real_video(body):
|
|
for ifr in body.find_all('iframe'):
|
|
if VID.search(ifr.get('src', '') or ''):
|
|
return True
|
|
for a in body.find_all('a', href=True):
|
|
if VID.search(a['href']):
|
|
return True
|
|
for v in body.find_all('video'):
|
|
src = v.get('src') or (v.find('source').get('src') if v.find('source') else '')
|
|
if src and not src.lower().endswith(('.gif', '.png', '.jpg')):
|
|
return True
|
|
return False
|
|
|
|
|
|
def real_audio(body):
|
|
for au in body.find_all('audio'):
|
|
if au.get('src') or au.find('source'):
|
|
return True
|
|
for a in body.find_all('a', href=True):
|
|
if AUD_EXT.search(a['href']):
|
|
return True
|
|
for s in body.find_all('source'):
|
|
if AUD_EXT.search(s.get('src', '') or ''):
|
|
return True
|
|
return False
|
|
|
|
|
|
def rebuild_N(N, vid, aud):
|
|
parts = []
|
|
for p in (N or '').split(','):
|
|
if p == '영상' and not vid:
|
|
continue
|
|
if p == '오디오' and not aud:
|
|
continue
|
|
if p:
|
|
parts.append(p)
|
|
return ','.join(parts) if parts else '없음'
|
|
|
|
|
|
def run(name, sess):
|
|
xlsx = os.path.join(OUTDIR, f'{name}.xlsx')
|
|
wb = openpyxl.load_workbook(xlsx)
|
|
ws = wb.active
|
|
chg = 0
|
|
for r in range(3, ws.max_row + 1):
|
|
if ws.cell(r, 2).value is None:
|
|
break
|
|
L = ws.cell(r, 12).value
|
|
N = ws.cell(r, 14).value or ''
|
|
if L == '사이트' or ('영상' not in N and '오디오' not in N):
|
|
continue
|
|
url = ws.cell(r, 11).value
|
|
if not url or not isinstance(url, str) or not url.startswith('http'):
|
|
continue
|
|
soup, _ = P.fetch(sess, url)
|
|
if not soup:
|
|
continue
|
|
body = P.get_body(soup, P.BODY_SEL)
|
|
vid = real_video(body); aud = real_audio(body)
|
|
if L == '게시판' and (('영상' in N and not vid) or ('오디오' in N and not aud)):
|
|
for du in P.extract_detail_urls(body, url, limit=4):
|
|
ds, _ = P.fetch(sess, du)
|
|
if not ds:
|
|
continue
|
|
db = P.get_body(ds, P.BODY_SEL)
|
|
vid = vid or real_video(db); aud = aud or real_audio(db)
|
|
if vid and aud:
|
|
break
|
|
newN = rebuild_N(N, vid, aud)
|
|
if newN != N:
|
|
ws.cell(r, 14).value = newN; chg += 1
|
|
wb.save(xlsx)
|
|
print(f'[{name}] 영상/오디오 거짓제거 {chg}행')
|
|
return chg
|
|
|
|
|
|
def main():
|
|
probe = {r['name']: r for r in json.load(open(r'D:\01.프로젝트\DB수집\_스크립트\_공공기관_probe.json', encoding='utf-8'))}
|
|
order = sorted(probe.values(), key=lambda x: -int(x['num']))
|
|
only = sys.argv[1:]
|
|
if only:
|
|
order = [p for p in order if p['name'] in only or str(p['num']) in only]
|
|
sess = P.make_session()
|
|
tot = 0
|
|
for p in order:
|
|
try:
|
|
tot += run(p['name'], sess)
|
|
except Exception as e:
|
|
print(f"[{p['name']}] 실패: {e}")
|
|
print(f'=== 영상/오디오 교정 합계 {tot}행 ===')
|
|
|
|
|
|
if __name__ == '__main__':
|
|
main()
|