名稱拆分和比較

Question

我有一個情況我需要做到以下幾點：

公司名稱：

a. Split text before and after 「 - 「 
b. Generate the report where texts before and after 「 - 「 matches = exact match 
c. Generate the report where texts before and after 「 - 「 matches = similar matches 

我能到達，直到B點。其中，我能得到具有相同firsthalf和secondhalf結果（如ABC，INC。 - ABC，INC。）名稱中使用下列 -

RTRIM(substring(c.companyname,0,charindex('-',c.companyname)))= LTRIM(substring(c.companyname, charindex('-',c.companyname,0)+1, len(c.companyname))) 

但是，我不能做下一個報告（如ABC - abc或abc，inc - abc）

有人可以幫忙嗎？

Answer 1

試試這個嗎？

DECLARE @CompanyNames TABLE (
    CompanyName VARCHAR(512)); 
INSERT INTO @CompanyNames VALUES ('Walt Disney - Walt Disney'); 
INSERT INTO @CompanyNames VALUES ('Fun Food - Fun Food'); 
INSERT INTO @CompanyNames VALUES ('Fun Food, Inc. - Fun Food'); 
INSERT INTO @CompanyNames VALUES ('Walt Disney - Walt Disney, Inc.'); 

--Split names 
DECLARE @SplitNames TABLE (
    MatchLeft VARCHAR(128), 
    MatchRight VARCHAR(128)); 
INSERT INTO 
    @SplitNames 
SELECT 
    RTRIM(SUBSTRING(CompanyName, 0, CHARINDEX('-', CompanyName))), 
    LTRIM(SUBSTRING(CompanyName, CHARINDEX('-', CompanyName, 0) + 1, LEN(CompanyName))) 
FROM 
    @CompanyNames; 

--Exact matches 
SELECT 
    MatchLeft, 
    MatchRight, 
    CASE WHEN MatchLeft = MatchRight THEN 1 ELSE 0 END AS Exact 
FROM 
    @SplitNames; 

--Inexact matches 
WITH CleansedCompanyNames AS (
    SELECT 
     MatchLeft AS OriginalMatchLeft, 
     MatchRight AS OriginalMatchRight, 
     REPLACE(REPLACE(REPLACE(MatchLeft, '.', ''), 'Inc', ''), ',', '') AS MatchLeft, 
     REPLACE(REPLACE(REPLACE(MatchRight, '.', ''), 'Inc', ''), ',', '') AS MatchRight 
    FROM 
     @SplitNames) 
SELECT 
    OriginalMatchLeft, 
    OriginalMatchRight, 
    MatchLeft, 
    MatchRight, 
    CASE WHEN MatchLeft = MatchRight THEN 1 ELSE 0 END 
FROM 
    CleansedCompanyNames; 

--Using SOUNDEX 
SELECT 
    MatchLeft, 
    MatchRight, 
    CASE WHEN DIFFERENCE(MatchLeft, MatchRight) >= 3 THEN 1 ELSE 0 END AS Score 
FROM 
    @SplitNames; 

有兩種思路有處理不精確匹配：

• 要麼匹配之前刪除標點和不受歡迎的詞彙（但這將需要建立什麼來代替列表）;或
• 使用SOUNDEX來測試字符串相似性。

或者，要使用你原來的例子，你可以用這個SOUNDEX：

SELECT ... 
WHERE 
DIFFERENCE(RTRIM(substring(c.companyname,0,charindex('-',c.companyname))), LTRIM(substring(c.companyname, charindex('-',c.companyname,0)+1, len(c.companyname)))) >= 3 

和你最新的例子：

DECLARE @Company TABLE (
    companyname VARCHAR(500)); 
INSERT INTO @Company VALUES ('Allen Limited - Allen Corporation'); 
INSERT INTO @Company VALUES ('Sweden Corp. - Sweden Corp.'); 
INSERT INTO @Company VALUES ('Alaska Limited - Alaska Limited, Inc.'); 
INSERT INTO @Company VALUES ('New York Inc. - New York Steel Limited'); 
INSERT INTO @Company VALUES ('India Plc - India Plc.'); 
INSERT INTO @Company VALUES ('Dubai International - Dubai International'); 
INSERT INTO @Company VALUES ('Nigera Falls Pvt. Ltd. - Amazing Nigeria Falls'); 
SELECT 
    c.companyname, 
    DIFFERENCE(RTRIM(SUBSTRING(c.companyname, 0, CHARINDEX('-', c.companyname))), LTRIM(SUBSTRING(c.companyname, CHARINDEX('-', c.companyname, 0) + 1, LEN(c.companyname)))) AS Similarity 
FROM 
    @Company c; 

有了結果：

companyname Similarity 
Allen Limited - Allen Corporation 4 
Sweden Corp. - Sweden Corp. 4 
Alaska Limited - Alaska Limited, Inc. 4 
New York Inc. - New York Steel Limited 4 
India Plc - India Plc. 4 
Dubai International - Dubai International 4 
Nigera Falls Pvt. Ltd. - Amazing Nigeria Falls 1 

所以對於你的最後一個例子來說效果不太好，但對於t來說似乎很好他人呢？