如何在模糊匹配后存储与名称最匹配的子串而非完整字符串

我正在用Python对字符串列表中的姓名进行模糊匹配，已生成一个字典——键为匹配到的姓名，值为该姓名所在的完整来源字符串。现在需要生成新字典，将值替换为与键中姓名最匹配的子串，例如把{'Volodymyr Zelenskyy': 'discussed fighter jets with President Volodymyr Zelensky'}转换为{'Volodymyr Zelenskyy': 'Volodymyr Zelensky'}。

现有实现代码 #

import pandas as pd
from fuzzywuzzy import process
from fuzzywuzzy import fuzz

master_list=['Muhammad bin Rashid Al Maktum','Antonio Costa','Nikol Pashinyan','Antony Blinken','Mohammad Shtayyeh','Mohammad Shtayyeh','Sebastian Kurz','Kyriakos Mitsotakis','Volodymyr Zelenskyy','Sebastian Kurz']
notes_list=df.Notes.tolist() 
expanded_notes = []
expanded_notes_2 = []
expanded_notes_full = []
matched_names = []
matching_strings = []
threshold=86

# 拆分笔记字符串，扩展为单个条目列表
for string in notes_list:
    s_list = string.strip("").split(';') 
    expanded_notes.extend(s_list)
for string in expanded_notes:
    s_list = string.strip("").split(',') 
    expanded_notes_2.extend(s_list)
for string in expanded_notes_2:
    s_list = string.strip("").split(' and ') 
    expanded_notes_full.extend(s_list)
    
# 模糊匹配姓名，生成初始字典
for string in expanded_notes_full:
    for name in master_list:
        if fuzz.partial_ratio(name, string) >= threshold:
            matched_names.append(name)
            matching_strings.append(string)
                                       
d= dict(zip(matched_names, matching_strings))

示例字典 #

d={'Muhammad bin Rashid Al Maktum': 'Met with Prime Minister Muhammad bin Rashid Al Maktum',
 'Antonio Costa': 'Met with Prime Minister Antonio Costa',
 'Nikol Pashinyan': 'met with Prime Minister Nikol Pashinyan',
 'Antony Blinken': 'Secretary of State Antony Blinken',
 'Mohammad Shtayyeh': 'Met with Prime Minister Mohammad Shtayyeh',
 'Kyriakos Mitsotakis': 'Met with Prime Minister Kyriakos Mitsotakis',
 'Sebastian Kurz': 'Sebastian Kurz',
 'Volodymyr Zelenskyy': 'discussed fighter jets with President Volodymyr Zelensky'}

解决方案 #

通过遍历初始字典的键值对，针对每个值字符串生成所有可能的连续单词子串，再用模糊匹配找到与键姓名匹配度最高的子串，即可生成目标字典：

实现代码 #

from fuzzywuzzy import process, fuzz

def generate_substring_candidates(text):
    """生成文本中所有连续单词组成的子串候选"""
    words = text.split()
    candidates = []
    for start in range(len(words)):
        for end in range(start + 1, len(words) + 1):
            candidates.append(' '.join(words[start:end]))
    return candidates

# 生成新字典
new_matched_dict = {}
for target_name, source_text in d.items():
    candidates = generate_substring_candidates(source_text)
    # 提取匹配度最高的子串
    best_match, _ = process.extractOne(target_name, candidates, scorer=fuzz.partial_ratio)
    new_matched_dict[target_name] = best_match

效果验证 #

运行上述代码后，示例字典会被转换为：

{
 'Muhammad bin Rashid Al Maktum': 'Muhammad bin Rashid Al Maktum',
 'Antonio Costa': 'Antonio Costa',
 'Nikol Pashinyan': 'Nikol Pashinyan',
 'Antony Blinken': 'Antony Blinken',
 'Mohammad Shtayyeh': 'Mohammad Shtayyeh',
 'Kyriakos Mitsotakis': 'Kyriakos Mitsotakis',
 'Sebastian Kurz': 'Sebastian Kurz',
 'Volodymyr Zelenskyy': 'Volodymyr Zelensky'
}

内容的提问来源于stack exchange，提问作者LennyAngola