python_ai/文档一二级标题识别与提取.py

import re
import time
from docx import Document
from pprint import pprint
# from paddlenlp import Taskflow
#
# similarity = Taskflow("text_similarity", truncation=True, max_length=102400)


def getOutlineLevel(inputXml):
    """
    功能 从xml字段中提取出<w:outlineLvl w:val="number"/>中的数字number
    参数 inputXml
    返回 number
    """
    start_index = inputXml.find('<w:outlineLvl')
    end_index = inputXml.find('>', start_index)
    number = inputXml[start_index:end_index + 1]
    number = re.search("\d+", number).group()
    return number


def isTitle(paragraph):
    """
    功能 判断该段落是否设置了大纲等级
    参数 paragraph:段落
    返回 None:普通正文，没有大纲级别 0:一级标题 1:二级标题 2:三级标题
    """
    # 如果是空行，直接返回None
    if paragraph.text.strip() == '':
        return None

    # 如果该段落是直接在段落里设置大纲级别的，根据xml判断大纲级别
    paragraphXml = paragraph._p.xml
    if paragraphXml.find('<w:outlineLvl') >= 0:
        return getOutlineLevel(paragraphXml)
    # 如果该段落是通过样式设置大纲级别的，逐级检索样式及其父样式，判断大纲级别
    targetStyle = paragraph.style
    while targetStyle is not None:
        # 如果在该级style中找到了大纲级别，返回
        if targetStyle.element.xml.find('<w:outlineLvl') >= 0:
            return getOutlineLevel(targetStyle.element.xml)
        else:
            targetStyle = targetStyle.base_style
    # 如果在段落、样式里都没有找到大纲级别，返回None
    return None

def getDocxToText12biaoti(name):
    document = Document(name)
    # 逐段读取docx文档的内容
    levelList=[]
    words=[]
    levelText=""
    i = 0
    firstTitle = 0
    secondTitle = 0
    sanjiTitle = 0
    for paragraph in document.paragraphs:
        # 判断该段落的标题级别
        # 这里用isTitle()临时代表，具体见下文介绍的方法
        text = paragraph.text

        if text.strip():#非空判断
            # print("非空")
            # words.append(text)
            level = isTitle(paragraph)
            if level=="0":
                firstTitle+=1
                secondTitle = 0
                if(text.find("附件")>=0):
                    continue
                words.append("{}:".format(firstTitle)+text)
            elif level=="1":
                secondTitle+=1
                sanjiTitle=0
                # words.append("\t"+"{}.{}".format(firstTitle,secondTitle)+text)
                words.append("{}.{}".format(firstTitle,secondTitle)+text)
            elif level=="2":
                sanjiTitle += 1
                # words.append("\t"+"{}.{}".format(firstTitle,secondTitle)+text)
                words.append("{}.{}.{}".format(firstTitle, secondTitle,sanjiTitle) + text)


    # 将所有段落文本拼接成一个字符串，并用换行符分隔
    print(len(words))
    if len(words)==0:
        raise Exception("I know python!")
    text = '\n'.join(words)
    with open("ce1.txt", 'w',encoding="utf-8") as txt_file:
        txt_file.write(text)
    return words
mobanList=[]
dangqianList=[]
errorList =[]
    # 将文本写入txt文件
# with open("ce模板.txt", 'r',encoding="utf-8") as txt_file:
#     for i in txt_file:
#         i=re.sub(r'[\t\n]', '', i)
#         mobanList.append(i)
# pprint(mobanList)
# dangqianList=getDocxToText12biaoti("1.docx")
# if len(dangqianList)!=len(mobanList):
#     print("标题数量与模板不一致")
# for num in range(len(mobanList)):
#     moban = mobanList[num]
#     dangqian= dangqianList[num]
#     fenshu=similarity([[dangqian,moban]])
#     pprint(fenshu)
#     if (fenshu[0]["similarity"]<0.85):
#         errorList.append(dangqianList)
# getDocxToText12biaoti("1.docx")
# pprint(errorList)

prompt = '''{}这是文档大纲，根据大纲分析文档中是否有{}这块内容的描述,若不存在请回答不存在
'''
dagang ="1"
biaozhun="2"
print(prompt.format(dagang, biaozhun))