如何实现列表间前缀模糊匹配并区分匹配与未找到项

需求说明 #

对to_search列表中的每个值，在choices列表中执行严格前缀模糊匹配（仅认可前缀顺序完全一致的匹配），返回result列表中对应的项；无匹配项时显示NOT FOUND。

原代码问题 #

原代码使用rapidfuzz的process.extractOne和fuzz.WRatio，存在两个核心问题：

• 误匹配：例如38050被错误匹配到358，不符合前缀匹配要求
• 未匹配项标识错误：例如49185、51078错误匹配到1，未返回NOT FOUND

原代码：

from rapidfuzz import process, fuzz

choices = [
        '237','1721','124622','334','124624','124','1246','1876','358',
        '33751','33679','599','61','230','31','65','1721','1','124623'
    ]

result = [
            'NAD','ATE','STA','SSI','GYP','RIC','EEC','AND','GIU','ANC',
            'PAI','GAR','TAL','ANI','LAN','TRI','GDO','MAR','EDE'
        ]

to_search = ['18763044','187635','23092','3162','38050','33','49185','51078','1246','1721']

for element in to_search:
    match =  process.extractOne(element, choices, scorer=fuzz.WRatio)
    print(element,result[match[2]],'         ## ',match)

当前错误输出 #

18763044    AND         ##  ('1876', 90.0, 7)
187635      AND         ##  ('1876', 90.0, 7)
23092       ANI         ##  ('230', 90.0, 13)
3162        LAN         ##  ('31', 90.0, 14)
38050       GIU         ##  ('358', 72.0, 8) // 应标记为NOT FOUND
33          SSI         ##  ('334', 90.0, 3)
49185       MAR         ##  ('1', 90.0, 17)  // 应标记为NOT FOUND
51078       MAR         ##  ('1', 90.0, 17)  // 应标记为NOT FOUND
1246        EEC         ##  ('1246', 100.0, 6)
1721        ATE         ##  ('1721', 100.0, 1)

解决方案 #

要实现严格前缀匹配，需自定义评分规则并设置匹配阈值：

1. 自定义评分函数，仅当choice是element的前缀时计算有效分数，否则返回0
2. 设置分数阈值，低于阈值时判定为无匹配

修改后的代码：

from rapidfuzz import process, fuzz

choices = [
        '237','1721','124622','334','124624','124','1246','1876','358',
        '33751','33679','599','61','230','31','65','1721','1','124623'
    ]

result = [
            'NAD','ATE','STA','SSI','GYP','RIC','EEC','AND','GIU','ANC',
            'PAI','GAR','TAL','ANI','LAN','TRI','GDO','MAR','EDE'
        ]

to_search = ['18763044','187635','23092','3162','38050','33','49185','51078','1246','1721']

def prefix_scorer(query, choice, **kwargs):
    # 仅当choice是query的前缀时计算分数，否则返回0
    if query.startswith(choice):
        return fuzz.WRatio(query, choice)
    return 0

for element in to_search:
    match = process.extractOne(element, choices, scorer=prefix_scorer, score_cutoff=80)
    if match:
        print(element, result[match[2]])
    else:
        print(element, "NOT FOUND")

代码解释 #

• prefix_scorer函数：从根源上过滤非前缀匹配，仅当choice是query的前缀时才计算匹配分数
• score_cutoff=80：过滤低质量匹配，无有效匹配时match为None，此时输出NOT FOUND

最终输出 #

18763044    AND
187635      AND
23092       ANI
3162        LAN
38050       NOT FOUND
33          SSI
49185       NOT FOUND
51078       NOT FOUND
1246        EEC
1721        ATE

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