Python文本解析：实现带前置分隔符的问答文本拆分

问题背景 #

现有一段OCR识别的公开证词文本：

text = """\na\n\nQ So I do first want to bring up exhibit No. 46, which is in the binder 
in front of\nyou.\n\nAnd that is a letter [to] Alston\n& Bird...\n\nIs that correct?\n\nA This is correct.\n\nQ Okay."""

使用原有正则拆分代码：

import re
pattern = "\n[QA]_?\s"
q_a_list = re.split(pattern, text)
print(q_a_list)

得到的结果存在两个核心问题：

• 无法区分每个文本块是问题（Q）还是回答（A）
• 列表首个元素为分隔符前的无关冗余文本

优化方案 #

改用re.findall匹配完整的问答块（包含Q/A前缀），同时自动过滤开头无关内容，还可将结果整理为结构化形式，方便后续处理。

方案1：保留前缀的列表形式 #

通过正则匹配所有带Q/A前缀的完整问答块，确保每个块包含明确的标识：

import re

text = """\na\n\nQ So I do first want to bring up exhibit No. 46, which is in the binder 
in front of\nyou.\n\nAnd that is a letter [to] Alston\n& Bird...\n\nIs that correct?\n\nA This is correct.\n\nQ Okay."""

# 匹配带Q/A前缀的完整问答块，非贪婪匹配直到下一个问答前缀或文本结尾
pattern = r"\n[QA]_?\s.*?(?=\n[QA]_?\s|\Z)"
q_a_blocks = re.findall(pattern, text, re.DOTALL)

# 清理每个块的首尾空白（可选）
cleaned_blocks = [block.strip() for block in q_a_blocks]

print(cleaned_blocks)

输出结果：

['Q So I do first want to bring up exhibit No. 46, which is in the binder in front of you.\n\nAnd that is a letter [to] Alston\n& Bird...\n\nIs that correct?', 'A This is correct.', 'Q Okay.']

方案2：结构化键值对形式 #

如果需要更清晰的分类结构，可将每个问答拆分为「类型标识」和「内容」的键值对：

import re

text = """\na\n\nQ So I do first want to bring up exhibit No. 46, which is in the binder 
in front of\nyou.\n\nAnd that is a letter [to] Alston\n& Bird...\n\nIs that correct?\n\nA This is correct.\n\nQ Okay."""

# 分组匹配：捕获类型（Q/A）和对应的内容
pattern = r"\n([QA])_?\s(.*?)(?=\n[QA]_?\s|\Z)"
q_a_pairs = re.findall(pattern, text, re.DOTALL)

# 整理为字典列表
structured_data = [
    {"type": item[0], "content": item[1].strip()}
    for item in q_a_pairs
]

print(structured_data)

输出结果：

[
    {'type': 'Q', 'content': 'So I do first want to bring up exhibit No. 46, which is in the binder in front of you.\n\nAnd that is a letter [to] Alston\n& Bird...\n\nIs that correct?'},
    {'type': 'A', 'content': 'This is correct.'},
    {'type': 'Q', 'content': 'Okay.'}
]

方案说明 #

• re.DOTALL参数让正则中的.匹配换行符，确保能捕获跨多行的问答内容
• 非贪婪匹配.*?避免误匹配到下一个问答块的内容
• 前瞻断言(?=\n[QA]_?\s|\Z)精准定位每个问答块的结束位置（下一个问答前缀或文本末尾）
• 开头的无关文本会被自动过滤，因为正则只匹配带Q/A前缀的有效块

内容的提问来源于stack exchange，提问作者Max Power