[Python] 纯文本查看 复制代码
import os
import re
import sys
import configparser
import spacy
from docx import Document
import win32com.client as win32
# ================= 1. 读取配置文件 =================
def load_config():
config = configparser.ConfigParser()
config.optionxform = str # 强制保留键的大小写，防止大写字母变小写
# 确定配置文件的路径 (兼容打包成 EXE 后的运行路径)
if getattr(sys, 'frozen', False):
application_path = os.path.dirname(sys.executable)
else:
application_path = os.path.dirname(os.path.abspath(__file__))
config_file = os.path.join(application_path, "doc_to_TM.ini")
if not os.path.exists(config_file):
print(f"【错误】找不到配置文件: {config_file}")
print("请确保 config.ini 与本程序放在同一文件夹下！")
os.system("pause")
sys.exit(1)
# 读取文件，使用 utf-8-sig 兼容带 BOM 的 UTF-8 编码
config.read(config_file, encoding='utf-8-sig')
# 安全地读取 LogoScanNum（如果没填或填了非数字，默认用20）
try:
logo_scan_num = int(config.get("Settings", "LogoScanNum", fallback=20))
except ValueError:
print("【警告】LogoScanNum 配置不正确，已自动重置为默认值 20")
logo_scan_num = 20
# 解析配置
settings = {
"NAME_PLACEHOLDER": config.get("Settings", "NamePlaceholder", fallback="某某"),
"ORG_PLACEHOLDER": config.get("Settings", "OrgPlaceholder", fallback="某单位"),
"LOGO_SCAN_NUM": logo_scan_num
}
whitelist_str = config.get("WhiteList", "Items", fallback="")
whitelist = [item.strip() for item in whitelist_str.split(",") if item.strip()]
replacements = dict(config.items("Replacements")) if config.has_section("Replacements") else {}
return settings, whitelist, replacements
# 加载配置
SETTINGS, WHITE_LIST, CUSTOM_REPLACEMENTS = load_config()
# ================= 2. 加载 AI 模型 =================
print("正在加载 AI 模型，请稍候...")
try:
nlp = spacy.load("zh_core_web_sm")
except Exception as e:
print(f"模型加载失败: {e}")
print("【错误】缺少 zh_core_web_sm 模型！")
os.system("pause")
sys.exit(1)
# ================= 3. 核心处理逻辑 =================
def smart_anonymize(text):
if not text or not text.strip():
return text
# ====================================================
# 新增：第一步 —— 开启“绝对保护罩”
# 将文本中出现的白名单词汇，临时替换成特殊占位符 (如 [[WP_0]])
# ====================================================
protected_items = {}
for i, word in enumerate(WHITE_LIST):
# 忽略大小写进行查找和保护 (比如保护 cpu 也会保护 CPU)
pattern = re.compile(re.escape(word), re.IGNORECASE)
# 查找文本中是否包含该白名单词
matches = set(pattern.findall(text))
for match_str in matches:
placeholder = f"[[WP_{i}]]"
protected_items[placeholder] = match_str
text = text.replace(match_str, placeholder)
# ====================================================
# 第二步 —— 正常的脱密逻辑 (此时白名单词已被隐藏，绝对安全)
# ====================================================
# 1. 关键词替换
sorted_keywords = sorted(CUSTOM_REPLACEMENTS.keys(), key=len, reverse=True)
for old in sorted_keywords:
text = text.replace(old, CUSTOM_REPLACEMENTS[old])
# 2. 正则匹配单位 (此时 "结构布局" 已经变成了 "结构布[[WP_x]]"，不会再被 "局" 误杀)
unit_pattern = r"([\u4e00-\u9fa5]{2,}(?:有限公司|股份公司|公司|研究所|集团|中心|委员会|项目组|厂|局))"
text = re.sub(unit_pattern, SETTINGS["ORG_PLACEHOLDER"], text)
# 3. AI 识别
doc = nlp(text)
entities = sorted(doc.ents, key=lambda x: len(x.text), reverse=True)
for ent in entities:
# 只要不是特殊占位符，就正常替换
if not ent.text.startswith("[[WP_"):
if ent.label_ == "PERSON":
text = text.replace(ent.text, SETTINGS["NAME_PLACEHOLDER"])
elif ent.label_ == "ORG" and ent.text not in CUSTOM_REPLACEMENTS.values():
text = text.replace(ent.text, SETTINGS["ORG_PLACEHOLDER"])
# ====================================================
# 新增：第三步 —— 解除保护，恢复白名单词汇
# ====================================================
for placeholder, original_word in protected_items.items():
text = text.replace(placeholder, original_word)
return text
def clean_images_via_word_engine(input_path):
print(f" -> 正在启动 Word 内核清理图片...")
word = win32.DispatchEx('Word.Application')
word.Visible = False
word.DisplayAlerts = 0
try:
doc = word.Documents.Open(os.path.abspath(input_path))
# 扫荡页眉页脚的所有图形和嵌入式图片
for section in doc.Sections:
for hf_type in [1, 2, 3]:
header = section.Headers(hf_type)
if header.Exists:
for i in range(header.Shapes.Count, 0, -1): header.Shapes(i).Delete()
for i in range(header.Range.InlineShapes.Count, 0, -1): header.Range.InlineShapes(i).Delete()
footer = section.Footers(hf_type)
if footer.Exists:
for i in range(footer.Shapes.Count, 0, -1): footer.Shapes(i).Delete()
for i in range(footer.Range.InlineShapes.Count, 0, -1): footer.Range.InlineShapes(i).Delete()
# 动态获取扫描段落数
scan_num = SETTINGS["LOGO_SCAN_NUM"]
check_limit = min(scan_num, doc.Paragraphs.Count)
if check_limit > 0:
limit_end = doc.Paragraphs(check_limit).Range.End
for i in range(1, check_limit + 1):
rng = doc.Paragraphs(i).Range
for j in range(rng.InlineShapes.Count, 0, -1): rng.InlineShapes(j).Delete()
for i in range(doc.Shapes.Count, 0, -1):
try:
if doc.Shapes(i).Anchor.Start <= limit_end:
doc.Shapes(i).Delete()
except Exception:
pass
temp_docx = os.path.abspath(input_path) + "_temp.docx"
doc.SaveAs(temp_docx, FileFormat=16)
doc.Saved = True
doc.Close()
return temp_docx
except Exception as e:
print(f" Word 内核处理失败: {e}")
raise e
finally:
word.Quit(0)
def process_document(input_file, output_file):
print(f"\n正在处理: {os.path.basename(input_file)}")
temp_working_file = clean_images_via_word_engine(input_file)
try:
print(f" -> 正在进行文字脱密...")
doc = Document(temp_working_file)
# ================== 替换这里的函数 ==================
def handle_paragraphs(paras):
for p in paras:
original_text = p.text
if not original_text.strip():
continue
new_text = smart_anonymize(original_text)
if new_text != original_text:
if p.runs:
p.runs[0].text = new_text
for i in range(1, len(p.runs)):
p.runs[i].text = ""
# ====================================================
handle_paragraphs(doc.paragraphs)
for table in doc.tables:
for row in table.rows:
for cell in row.cells:
handle_paragraphs(cell.paragraphs)
for section in doc.sections:
if section.header:
handle_paragraphs(section.header.paragraphs)
for table in section.header.tables:
for row in table.rows:
for cell in row.cells: handle_paragraphs(cell.paragraphs)
if section.footer:
handle_paragraphs(section.footer.paragraphs)
output_file_docx = output_file if output_file.endswith(".docx") else output_file + "x"
doc.save(output_file_docx)
print(f" √ 成功完成: {os.path.basename(output_file_docx)}")
finally:
if os.path.exists(temp_working_file):
os.remove(temp_working_file)
def batch_process_documents():
if getattr(sys, 'frozen', False):
current_dir = os.path.dirname(sys.executable)
else:
current_dir = os.getcwd()
extensions = ('.doc', '.docx')
print(f"====== 开始静默批量脱密处理 ======")
file_count = 0
for file_name in os.listdir(current_dir):
if (file_name.lower().endswith(extensions) and
not file_name.startswith("~$") and
not file_name.rsplit('.', 1)[0].endswith('_TM') and
not file_name.endswith('_temp.docx')):
input_path = os.path.join(current_dir, file_name)
name_part, extension = os.path.splitext(file_name)
output_name = f"{name_part}_TM{extension}"
output_path = os.path.join(current_dir, output_name)
try:
process_document(input_path, output_path)
file_count += 1
except Exception as e:
print(f" × 处理出错 {file_name}: {e}")
print("==============================")
print(f"所有文件处理完毕！共处理了 {file_count} 个文件。")
os.system("pause")
if __name__ == "__main__":
batch_process_documents()
发表于 2026-4-14 15:08
|
发表于 2026-4-14 15:27
|楼主
分享到朋友圈