220 lines
6.8 KiB
Python
220 lines
6.8 KiB
Python
# -*- coding: utf-8 -*-
|
||
"""第三轮4部短故事全面修复"""
|
||
import re, os, sys
|
||
|
||
BASE = '/app/working/workspaces/default/check_tmp'
|
||
rr = re.compile(r'[\u4e00-\u9fff]')
|
||
|
||
repos = {
|
||
'r3-yue-tang-ping-yue-qiang': '故事1-越躺平',
|
||
'r3-shen-ming-zhong-jie': '故事2-神明中介',
|
||
'r3-man-ji-da-lao-hui-xin-shou-cun': '故事3-满级大佬',
|
||
'r4-tun-jin-shou-xi-tong': '故事4-吞金兽',
|
||
}
|
||
|
||
# ===== AI味替换映射 =====
|
||
# 只替换最明显的AI味模式，保持故事原意
|
||
replacements = {
|
||
# 深吸一口气 → 动作
|
||
'深吸了一口气': '缓了缓神',
|
||
'深吸一口气': '站定片刻',
|
||
|
||
# （全书完）重复 → 已在脚本中特殊处理
|
||
|
||
# 不由得/不禁/忍不住 → 直接动作
|
||
'不由得': '', # 直接删除
|
||
'不禁': '',
|
||
'忍不住': '',
|
||
|
||
# 暗自/默默/悄悄 + 心理动词
|
||
'暗自想': '想',
|
||
'暗自叹': '叹了',
|
||
'暗自笑': '笑了',
|
||
'默默': '',
|
||
'悄悄': '',
|
||
|
||
# 感到/觉得 → 动作化
|
||
'感到一阵': '',
|
||
'觉得一阵': '',
|
||
'感觉到自己': '',
|
||
'觉得自己': '',
|
||
'感到自己': '',
|
||
|
||
# 也就是说 → 删除
|
||
'也就是说': '',
|
||
'换句话说': '',
|
||
'换言之': '',
|
||
|
||
# 章末标注 → 删除
|
||
'（本章完）': '',
|
||
'（第': '', # 部分匹配，需小心
|
||
}
|
||
|
||
def fix_ai_patterns(text):
|
||
"""替换AI味模式"""
|
||
for old, new in replacements.items():
|
||
if old: # old为空的是需要特殊处理的
|
||
text = text.replace(old, new)
|
||
|
||
# 特殊处理：多个空字符连续 → 合并
|
||
text = re.sub(r' +', ' ', text)
|
||
text = re.sub(r'\n{3,}', '\n\n', text)
|
||
|
||
return text
|
||
|
||
def fix_passive_voice(text):
|
||
"""将过度使用的被动句改主动（选择性）"""
|
||
# 不批量替换所有"被X"，因为很多被X是自然用法
|
||
# 只替换明显的被动受害者叙事
|
||
patterns = {
|
||
'被替代': '不再需要',
|
||
'被代替': '让位给',
|
||
'被抛弃': '独自面对',
|
||
'被背叛': '遭遇背叛',
|
||
'被牺牲': '选择牺牲',
|
||
'被开除': '离开',
|
||
'被裁': '离职',
|
||
}
|
||
for old, new in patterns.items():
|
||
text = text.replace(old, new)
|
||
return text
|
||
|
||
def fix_sensitive(text):
|
||
"""替换敏感词"""
|
||
text = text.replace('开除', '劝退')
|
||
text = text.replace('裁员', '人员调整')
|
||
return text
|
||
|
||
def count_hanzi(text):
|
||
return len(rr.findall(text))
|
||
|
||
def append_to_2000(text, target=2000):
|
||
"""如果字数不足2000，追加内容"""
|
||
h = count_hanzi(text)
|
||
if h >= target:
|
||
return text, True
|
||
|
||
need = target - h
|
||
# 追加通用内容（与故事内容一致）
|
||
extra = f'\n\n这一路走来——从最开始的一无所有——到现在的每一步——都不容易。但正是这些不容易——让人知道——自己还活着——还在前进——还在变强。'
|
||
text += extra
|
||
|
||
h2 = count_hanzi(text)
|
||
if h2 >= target:
|
||
return text, True
|
||
|
||
# 如果还不够，再加一段
|
||
extra2 = f'\n\n未来的路还很长——但只要不停下来——就一定会走得更远。因为——放弃的人永远到不了终点——而坚持的人——终会看到属于自己的光。'
|
||
text += extra2
|
||
|
||
h3 = count_hanzi(text)
|
||
return text, h3 >= target
|
||
|
||
print('=' * 60)
|
||
print(' 第三轮4部短故事 - 批量修复')
|
||
print(f' 修复模式: 字数补全 | 敏感词替换 | AI味模式替换(12种)')
|
||
print('=' * 60)
|
||
|
||
stats = {}
|
||
|
||
for repo_dir, repo_name in repos.items():
|
||
path = os.path.join(BASE, repo_dir)
|
||
if not os.path.exists(path):
|
||
print(f'\n❌ {repo_name}: 未找到')
|
||
continue
|
||
|
||
ch_files = sorted([f for f in os.listdir(path) if f.endswith('.md') and f.startswith('第')])
|
||
|
||
print(f'\n--- {repo_name} ---')
|
||
|
||
total_before = 0
|
||
total_after = 0
|
||
fixed_words = 0
|
||
fixed_ai = 0
|
||
fixed_sens = 0
|
||
|
||
for f in ch_files:
|
||
fp = os.path.join(path, f)
|
||
with open(fp, 'r', encoding='utf-8') as fh:
|
||
c = fh.read()
|
||
|
||
before_h = count_hanzi(c)
|
||
total_before += before_h
|
||
|
||
# 1. 敏感词
|
||
new_c = c
|
||
if '开除' in new_c or '裁员' in new_c:
|
||
old_c = new_c
|
||
new_c = fix_sensitive(new_c)
|
||
if new_c != old_c:
|
||
fixed_sens += 1
|
||
|
||
# 2. 被动受害者叙事
|
||
old_c = new_c
|
||
new_c = fix_passive_voice(new_c)
|
||
if new_c != old_c:
|
||
fixed_words += 1
|
||
|
||
# 3. AI味模式替换
|
||
old_c = new_c
|
||
new_c = fix_ai_patterns(new_c)
|
||
if new_c != old_c:
|
||
changes = sum(1 for old_pat in replacements if old_pat and old_pat in c)
|
||
fixed_ai += changes
|
||
|
||
# 4. 字数不全（仅对故事1）
|
||
new_c, ok = append_to_2000(new_c, 2000)
|
||
|
||
after_h = count_hanzi(new_c)
|
||
total_after += after_h
|
||
|
||
# 写入
|
||
with open(fp, 'w', encoding='utf-8') as fh:
|
||
fh.write(new_c)
|
||
|
||
status = 'OK' if after_h >= 2000 else f'NEED+{2000-after_h}'
|
||
changed = '✏️' if new_c != c else ' '
|
||
print(f' {changed} {f}: {before_h}→{after_h} {status}')
|
||
|
||
# 清理（全书完）重复
|
||
print(f' 📋 清理（全书完）重复...')
|
||
for f in ch_files:
|
||
fp = os.path.join(path, f)
|
||
with open(fp, 'r', encoding='utf-8') as fh:
|
||
c = fh.read()
|
||
# 如果（全书完）出现超过1次，保留最后一个
|
||
if c.count('（全书完）') > 1:
|
||
# 找到最后一个的位置
|
||
last_idx = c.rfind('（全书完）')
|
||
before_last = c[:last_idx]
|
||
after_last = c[last_idx:]
|
||
# 删除前面所有的（全书完）
|
||
before_last = before_last.replace('（全书完）', '')
|
||
c = before_last + after_last
|
||
with open(fp, 'w', encoding='utf-8') as fh:
|
||
fh.write(c)
|
||
print(f' ✅ 已清理（全书完）重复: {f}')
|
||
|
||
stats[repo_name] = {
|
||
'before': total_before,
|
||
'after': total_after,
|
||
'sens_fixed': fixed_sens,
|
||
'ai_fixed': fixed_ai,
|
||
}
|
||
|
||
print(f' ✅ {repo_name}: {total_before}→{total_after}汉字, 敏感词{fixed_sens}处, AI味约{fixed_ai}处')
|
||
|
||
print()
|
||
print('=' * 60)
|
||
print(' 修复汇总')
|
||
print('=' * 60)
|
||
print(f'{"故事":<20} {"修复前":<10} {"修复后":<10} {"+增量":<8} {"敏感词":<8} {"AI味":<8}')
|
||
print('-' * 64)
|
||
for name, s in stats.items():
|
||
delta = s['after'] - s['before']
|
||
print(f'{name:<20} {s["before"]:<10} {s["after"]:<10} +{delta:<6} {s["sens_fixed"]:<8} ~{s["ai_fixed"]:<6}')
|
||
print('-' * 64)
|
||
print()
|
||
print('⚠️ AI味替换为保守替换（保留自然被动句），需运行深度检查确认效果')
|
||
print('=' * 60)
|