比较字符串相似性

在 SQL Server 数据库中清洗数据

Miriam Antona

Software Engineer

问题描述

脏乱的字符串

| airport_state |
|---------------|
| Caalifornia   |
| California    |
| Californiaa   |
| Illinois      |
| Ilynois       |
| Tejas         |
| Texas         |

SOUNDEX

DIFFERENCE

SOUNDEX

SOUNDEX ( character_expression )

语音学算法
返回 4 个字符的代码
基于英语规则，也适用于许多其他语言的词

SELECT SOUNDEX('Illinois') AS soundex_code1;
SELECT SOUNDEX('Ilynois') AS soundex_code2;
SELECT SOUNDEX('California') AS soundex_code3;

| soundex_code1 |   | soundex_code2 |   | soundex_code3 |
|---------------|   |---------------|   |---------------|
| I452          |   | I452          |   | I416          |

SOUNDEX - 工作原理

示例："Illinois"

写下单词首字母
首字母后将元音及字母"h""w""y"替为 0

"Illinois" -> I

"Illinois" -> Ill0n00s

SOUNDEX - 工作原理

示例："Illinois"

首字母后将辅音替换为数字

SOUNDEX - 工作原理

示例："Illinois"

首字母后将辅音替换为数字

字母	替换为
b, f, p, v	1
c, g, j, k, q, s, x, z	2
d, t	3
l	4
m, n	5
r	6

"Ill0n00s" -> I4405002

SOUNDEX - 工作原理

相邻相同数字合并为一个
移除所有 0
若字母对应数字与首位数字相同，则移除首位数字
代码少于 3 位时在末尾补 0
多于 3 位时截断为 3 位

"I4405002" -> I40502

"I40502" -> I452

"I452"（不再处理）

SOUNDEX - 例外情况

SELECT SOUNDEX('Arizona') AS soundex_code1;
SELECT SOUNDEX('Arkansas') AS soundex_code2;

| soundex_code1 |   | soundex_code2 |
|---------------|   |---------------|
| A625          |   | A625          |

SOUNDEX - 相似性检查

SELECT DISTINCT A1.airport_state
FROM airports A1 
INNER JOIN airports A2 
    ON SOUNDEX(A1.airport_state) = SOUNDEX(A2.airport_state)
    AND A1.airport_state <> A2.airport_state

| airport_state |
|---------------|
| Caalifornia   |
| California    |
| Californiaa   |
| Illinois      |
| Ilynois       |
| New Jersey    |
| New York      |
| Tejas         |
| Texas         |

SOUNDEX - 相似性检查

SELECT DISTINCT A1.airport_state
FROM airports A1 
INNER JOIN airports A2 
    ON SOUNDEX(REPLACE(A1.airport_state, ' ', '')) = SOUNDEX(REPLACE(A2.airport_state, ' ', ''))
    AND A1.airport_state <> A2.airport_state

"New York" -> "NewYork"

| airport_state |
|---------------|
| Caalifornia   |
| California    |
| Californiaa   |
| Illinois      |
| Ilynois       |
| Tejas         |
| Texas         |

DIFFERENCE

DIFFERENCE ( character_expression , character_expression )

比较两个 SOUNDEX 值
返回 0 到 4 的数值
- 0 -> 几乎无相似性
- 4 -> 非常相似或完全相同

DIFFERENCE

SELECT DIFFERENCE('Illinois', 'Ilynois') AS dif_1;

| dif1 |
|------|
|  4   |

SELECT DIFFERENCE('Illinois', 'California') AS dif_2;

| dif2 |
|------|
| 1    |

DIFFERENCE - 相似性检查

SELECT DISTINCT A1.airport_state, A2.airport_state
FROM airports A1 
INNER JOIN airports A2 
    ON DIFFERENCE(REPLACE(A1.airport_state, ' ', ''), REPLACE(A2.airport_state, ' ', '')) = 4
    AND A1.airport_state <> A2.airport_state

| airport_state | airport_state |
|---------------|---------------|
| Caalifornia   | California    |
| Caalifornia   | Californiaa   |
| California    | Caalifornia   |
| California    | Californiaa   |
| Californiaa   | Caalifornia   |
| Californiaa   | California    |
| Illinois      | Ilynois       |
| Ilynois       | Illinois      |
| Massachusetts | Michigan      |
| Tejas         | Texas         |
| Texas         | Tejas         |

Vamos praticar!

在 SQL Server 数据库中清洗数据