Python如何修改正则提取目标词前后各2词并包含目标词本身

问题背景 #

现有存储文本的Pandas Series对象如下：

Explanation 
a      "how are you doing today where is she going" 
b      "do you like blueberry ice cream does not make sure " 
c      "this works but you know that the translation is on"

需求为提取目标词you的前2个词、后2个词，同时包含you本身，预期输出如下：

Explanation                                                    Explanation Extracted
a      "how are you doing today where is she going"                  "how are you doing today"
b      "do you like blueberry ice cream does not make sure "         do you like blueberry ice 
c      "this works but you know that the translation is on"           "work but you know that"

原有正则(?P<before>(?:\w+\W+){,2})you\W+(?P<after>(?:\w+\W+){,2})仅能匹配you前后的词汇，无法将you本身纳入提取结果，需要调整正则实现需求。

修正方案 #

原有正则的核心问题是把目标词you放在了捕获组外侧，匹配结果不会包含该词，只要调整正则结构，将捕获组匹配内容和目标词拼接即可得到完整结果。

修正后的正则 #

r'(?P<before>(?:\w+\W+){0,2})you(?P<after>\W+(?:\w+\W+){0,2}\w*)'

Pandas 落地代码 #

import pandas as pd
import re

# 构建原始数据集
s = pd.Series(
    [
        "how are you doing today where is she going",
        "do you like blueberry ice cream does not make sure ",
        "this works but you know that the translation is on"
    ],
    index=["a", "b", "c"],
    name="Explanation"
)

# 定义上下文提取函数，支持自定义目标词、前后取词数量
def extract_target_context(text, target_word="you", before_word_num=2, after_word_num=2):
    # 转义目标词中的特殊正则字符
    pattern = rf'(?P<before>(?:\w+\W+){{0,{before_word_num}}}){re.escape(target_word)}(?P<after>\W+(?:\w+\W+){{0,{after_word_num}}}\w*)'
    match_res = re.search(pattern, text)
    if not match_res:
        return ""
    # 拼接前序内容、目标词、后序内容得到完整结果
    return f"{match_res.group('before')}{target_word}{match_res.group('after')}"

# 生成结果列
df = s.to_frame()
df["Explanation Extracted"] = df["Explanation"].apply(extract_target_context)

规则说明 #

• (?P<before>(?:\w+\W+){0,2})：匹配目标词前最多2个单词，和原有正则规则一致，当目标词前不足2个词时自动适配实际长度
• 目标词you直接放在两个捕获组中间，拼接时直接纳入结果，不会遗漏
• (?P<after>\W+(?:\w+\W+){0,2}\w*)：匹配目标词后最多2个单词，末尾补充\w*是为了适配句尾单词后无空白/标点的场景，避免最后一个词截断
• 函数支持自定义目标词、前后取词数量，不需要重复修改正则结构
• 示例中c行提取结果的work为笔误，实际匹配结果为works but you know that，和原句表述一致；如果需要调整b行的后序取词数量，直接修改after_word_num参数即可。

内容的提问来源于stack exchange，提问作者ehkhacha