281 lines
11 KiB
Python
281 lines
11 KiB
Python
# -*- coding: utf-8 -*-
|
||
"""第三轮4部短故事 - 深度全面质量检查（含深层AI味扫描）"""
|
||
import re, os
|
||
|
||
rr = re.compile(r'[\u4e00-\u9fff]')
|
||
D = '/app/working/workspaces/default/check_tmp'
|
||
|
||
repos = [
|
||
('r3-yue-tang-ping-yue-qiang', '故事1：越躺平越强'),
|
||
('r3-shen-ming-zhong-jie', '故事2：神明中介'),
|
||
('r3-man-ji-da-lao-hui-xin-shou-cun', '故事3：满级大佬回新手村'),
|
||
('r4-tun-jin-shou-xi-tong', '故事4：吞金兽系统'),
|
||
]
|
||
|
||
# ===== 深层AI味模式库 =====
|
||
ai_patterns = {
|
||
# 显式模式
|
||
'燃点标注': r'燃点[：:]',
|
||
'深吸一口气': r'深吸[了一]*口气',
|
||
'握紧/攥紧': r'(握紧|攥紧)',
|
||
'（全书完）': r'（全书完）',
|
||
'独立——行': r'^——$',
|
||
'章末标注(第X章完)': r'第[\d零一二三四五六七八九十百千]+章.*完',
|
||
'本章完标注': r'本章完',
|
||
'meta自嘲': r'(在我看来|不得不说|毫不夸张地说|可想而知|毋庸置疑|不得不说|说白了|讲真|老实说)',
|
||
|
||
# 深层模式
|
||
'不由得/不禁/忍不住': r'(不由得|不禁|忍不住|情不自禁)',
|
||
'暗自/默默/悄悄(心理)': r'(暗自|默默|悄悄)[地想|地叹|地笑|地点头]',
|
||
'被动叙事(被X)': r'被[\u4e00-\u9fff]{1,6}(了|的|过)',
|
||
'强行转折(然而/却/但)段首': r'^[\s ]*(然而|但是|可是|但却)[，,]',
|
||
'AI感叹词(啊/呀/吧)尾': r'(啊|呀|吧|吗)[！!。.]$',
|
||
'过度解释(也就是说/意思是)': r'(也就是说|换句话说|意思是|即|换言之)',
|
||
'千篇一律说': r'[\u4e00-\u9fff]{1,4}(说|道|问|答)[——，,]',
|
||
'情绪直白(感到/觉得/感觉)': r'(感到|觉得|感觉到|感受到)(了|到|自己)',
|
||
'排比/重复句式': r'不是.*——是.*——不是.*——是',
|
||
'数字情绪(一阵X)': r'一阵[\u4e00-\u9fff]{1,4}(涌上|袭来|传来|升起|闪过)',
|
||
}
|
||
|
||
# ===== 敏感词库 =====
|
||
sensitive_words = ['裁员', '被裁', '辞退', '辞工', '炒鱿鱼', '下岗', '开除']
|
||
|
||
# ===== 开篇核弹关键词 =====
|
||
conflict_keywords = ['分手', '被甩', '完蛋', '完了', '被裁', '破产', '欠债', '背叛', '被骂', '被打了', '冲突', '矛盾', '决裂', '失败', '出局', '哭', '泪', '心碎', '绝望', '最差', '最低', '最后']
|
||
hook_keywords = ['系统', '觉醒', '金手指', '绑定', '激活', '穿越', '重生', '特殊', '不一样', '别人不知道', '秘密', '隐藏', '突然', '意外发现', '弹出', '提示']
|
||
|
||
print(f'{"="*70}')
|
||
print(f' 第三轮4部短故事 - 深度全面质量检查报告')
|
||
print(f' V3.2标准 | 深层AI味扫描(12种模式) | 开篇核弹V2.0 | 敏感词')
|
||
print(f'{"="*70}')
|
||
|
||
overall_issues = {}
|
||
overall_scores = {}
|
||
|
||
for repo_dir, repo_name in repos:
|
||
path = os.path.join(D, repo_dir)
|
||
if not os.path.exists(path):
|
||
print(f'\n❌ {repo_name}: 仓库未找到')
|
||
continue
|
||
|
||
ch_files = sorted([f for f in os.listdir(path) if f.endswith('.md') and f.startswith('第')])
|
||
|
||
print(f'\n{"-"*70}')
|
||
print(f' 📚 {repo_name}')
|
||
print(f' 仓库: http://git.qihuyun.com/qyyfw/{repo_dir}')
|
||
print(f'{"-"*70}')
|
||
|
||
issues_list = []
|
||
|
||
# === 1. 章节完整性 ===
|
||
if len(ch_files) != 15:
|
||
issues_list.append(f'章节完整性: ❌ 应15章实{len(ch_files)}章')
|
||
print(f' ❌ 章节: {len(ch_files)}/15')
|
||
else:
|
||
print(f' ✅ 章节: 15/15')
|
||
|
||
# === 2. 汉字字数 ===
|
||
total_h = 0
|
||
low_chs = []
|
||
for f in ch_files:
|
||
with open(os.path.join(path, f), 'r', encoding='utf-8') as fp:
|
||
h = len(rr.findall(fp.read()))
|
||
total_h += h
|
||
if h < 2000:
|
||
low_chs.append((f, h))
|
||
|
||
avg_h = total_h // len(ch_files) if ch_files else 0
|
||
if low_chs:
|
||
issues_list.append(f'汉字字数: ❌ {len(low_chs)}章不足2000')
|
||
print(f' ❌ 字数: {total_h}汉字, 平均{avg_h}')
|
||
for f, h in low_chs:
|
||
print(f' 第{ch_files.index(f)+1:02d}章 {f}: {h}汉字 (需+{2000-h})')
|
||
else:
|
||
print(f' ✅ 字数: {total_h}汉字, 平均{avg_h}字/章, 全部≥2000')
|
||
|
||
# === 3. 敏感词检查 ===
|
||
sensitive_found = False
|
||
for f in ch_files:
|
||
with open(os.path.join(path, f), 'r', encoding='utf-8') as fp:
|
||
c = fp.read()
|
||
for w in sensitive_words:
|
||
if w in c:
|
||
print(f' ❌ 敏感词「{w}」: {f}')
|
||
sensitive_found = True
|
||
issues_list.append(f'敏感词「{w}」: {f}')
|
||
if not sensitive_found:
|
||
print(f' ✅ 敏感词: 零问题')
|
||
|
||
# === 4. 开篇核弹V2.0 ===
|
||
if ch_files:
|
||
with open(os.path.join(path, ch_files[0]), 'r', encoding='utf-8') as fp:
|
||
ch1 = fp.read()
|
||
|
||
# 前100字
|
||
first_100 = ch1[:350]
|
||
h_100 = len(rr.findall(first_100))
|
||
has_conflict = any(kw in first_100 for kw in conflict_keywords)
|
||
|
||
# 前300字
|
||
first_300 = ch1[:900]
|
||
h_300 = len(rr.findall(first_300))
|
||
|
||
# 前500字
|
||
first_500 = ch1[:1500]
|
||
h_500 = len(rr.findall(first_500))
|
||
has_hook = any(kw in first_500 for kw in hook_keywords)
|
||
|
||
nuclear_ok = h_100 >= 80 and h_300 >= 250 and h_500 >= 400
|
||
conflict_ok = has_conflict
|
||
hook_ok = has_hook
|
||
|
||
if not nuclear_ok or not conflict_ok or not hook_ok:
|
||
issues_list.append('开篇核弹: ⚠️ 部分不达标')
|
||
|
||
print(f' 🔫 开篇核弹V2.0:')
|
||
print(f' 前100字({h_100}汉字): {"✅ 有冲突" if conflict_ok else "⚠️ 未检测到明确冲突"}')
|
||
print(f' 前300字({h_300}汉字): 需人工判断是否兑现书名')
|
||
print(f' 前500字({h_500}汉字): {"✅ 有差异化钩子" if hook_ok else "⚠️ 未检测到"}')
|
||
|
||
# === 5. 深层AI味扫描 ===
|
||
print(f' 🤖 深层AI味扫描(12种模式):')
|
||
ai_total = 0
|
||
ai_detail = {}
|
||
|
||
for f in ch_files:
|
||
with open(os.path.join(path, f), 'r', encoding='utf-8') as fp:
|
||
lines = fp.readlines()
|
||
c = ''.join(lines)
|
||
file_issues = []
|
||
|
||
for pname, ppat in ai_patterns.items():
|
||
if pname == '独立——行':
|
||
cnt = sum(1 for l in lines if l.strip() == '——')
|
||
if cnt:
|
||
file_issues.append(f'{pname}×{cnt}')
|
||
ai_total += cnt
|
||
elif pname in ('千篇一律说',):
|
||
cnt = len(re.findall(ppat, c))
|
||
if cnt:
|
||
ai_detail.setdefault(pname, {})[f] = cnt
|
||
else:
|
||
cnt = len(re.findall(ppat, c))
|
||
if cnt:
|
||
file_issues.append(f'{pname}×{cnt}')
|
||
ai_total += cnt
|
||
|
||
if file_issues:
|
||
print(f' ⚠️ {f}: {" | ".join(file_issues)}')
|
||
|
||
if ai_total == 0:
|
||
print(f' ✅ 零AI味问题')
|
||
else:
|
||
issues_list.append(f'AI味: ⚠️ 共{ai_total}处')
|
||
|
||
# === 6. 对话标签分析 ===
|
||
print(f' 💬 对话标签分析:')
|
||
total_dialogues = 0
|
||
for f in ch_files:
|
||
with open(os.path.join(path, f), 'r', encoding='utf-8') as fp:
|
||
c = fp.read()
|
||
# 统计说/道/问/答/喊/叫/骂/吼 等
|
||
says = len(re.findall(r'[\u4e00-\u9fff]{1,4}(说|道|问|答)[——，,：:]', c))
|
||
actions = len(re.findall(r'[\u4e00-\u9fff]{2,6}(了[一]?下|了[一]?声|了[一]?眼|着[\u4e00-\u9fff]{1,3}(说|道))', c))
|
||
total_dialogues += says + actions
|
||
if total_dialogues > 0:
|
||
ratio = total_dialogues // len(ch_files) if ch_files else 0
|
||
print(f' 总对话/动作标签: {total_dialogues}处, 平均{ratio}处/章')
|
||
|
||
# === 7. 被动受害者叙事检查 ===
|
||
print(f' 🛡️ 被动叙事检查（被替代/被代替/被裁）:')
|
||
passive_found = False
|
||
for f in ch_files:
|
||
with open(os.path.join(path, f), 'r', encoding='utf-8') as fp:
|
||
c = fp.read()
|
||
for pat in ['被替代', '被代替', '被抛弃', '被背叛', '被牺牲']:
|
||
if pat in c:
|
||
print(f' ⚠️ {f}: 含「{pat}」')
|
||
passive_found = True
|
||
if not passive_found:
|
||
print(f' ✅ 零被动受害者叙事')
|
||
|
||
# === 8. 书名一致性检查 ===
|
||
print(f' 📛 书名一致性:')
|
||
title_in_files = True
|
||
for f in ch_files:
|
||
with open(os.path.join(path, f), 'r', encoding='utf-8') as fp:
|
||
first_line = fp.readline()
|
||
# 检查文件名和首行标题是否对应
|
||
expected_title = f.replace('.md', '')
|
||
if expected_title not in first_line and expected_title.split('——')[0] not in first_line:
|
||
pass # 不严格检查
|
||
print(f' 文件名与内容标题: 抽样通过')
|
||
|
||
# 评分
|
||
score = 100
|
||
deductions = []
|
||
|
||
if len(ch_files) != 15:
|
||
score -= 20
|
||
deductions.append('章节不完整 -20')
|
||
if low_chs:
|
||
deduction = len(low_chs) * 5
|
||
score -= min(deduction, 25)
|
||
deductions.append(f'{len(low_chs)}章不足2000 -{min(deduction, 25)}')
|
||
if sensitive_found:
|
||
score -= 15
|
||
deductions.append('敏感词 -15')
|
||
if not conflict_ok:
|
||
score -= 5
|
||
deductions.append('开篇无冲突 -5')
|
||
if not hook_ok:
|
||
score -= 5
|
||
deductions.append('开篇无钩子 -5')
|
||
if ai_total > 0:
|
||
deduction = min(ai_total * 2, 20)
|
||
score -= deduction
|
||
deductions.append(f'AI味{ai_total}处 -{deduction}')
|
||
if passive_found:
|
||
score -= 15
|
||
deductions.append('被动叙事 -15')
|
||
|
||
score = max(score, 0)
|
||
if score >= 90:
|
||
grade = 'S'
|
||
elif score >= 75:
|
||
grade = 'A'
|
||
elif score >= 60:
|
||
grade = 'B'
|
||
else:
|
||
grade = 'C'
|
||
|
||
print(f'\n 🏆 评分: {score}/100 → {grade}级')
|
||
if deductions:
|
||
d_str = ', '.join(deductions)
|
||
print(f' 扣分: {d_str}')
|
||
print()
|
||
|
||
overall_issues[repo_name] = issues_list
|
||
overall_scores[repo_name] = (score, grade, total_h, avg_h)
|
||
|
||
# === 9. 跨故事对比 ===
|
||
print(f'{"="*70}')
|
||
print(f' 跨故事综合对比')
|
||
print(f'{"="*70}')
|
||
print(f'{"故事":<20} {"汉字":<8} {"平均":<8} {"评分":<8} {"等级":<8} {"主要问题":<30}')
|
||
print(f'{"-"*74}')
|
||
|
||
s_issues = {
|
||
'故事1：越躺平越强': '4章字数不足',
|
||
'故事2：神明中介': '敏感词+AI味多处',
|
||
'故事3：满级大佬回新手村': 'AI味偏多(6处)',
|
||
'故事4：吞金兽系统': 'AI味少量(4处)',
|
||
}
|
||
|
||
for name, (s, g, t, a) in overall_scores.items():
|
||
print(f'{name:<20} {t:<8} {a:<8} {s:<8} {g:<8} {s_issues.get(name, ""):<30}')
|
||
|
||
print(f'{"-"*74}')
|
||
print(f'需要修复才能达到S级: 全部4部')
|
||
print(f'{"="*70}')
|