如何从 Oracle table 获取几乎匹配的字符串？

Question

我在 Oracle 中有一个包含四列的 table。

现在用户可以将输入字符串作为 "operation Knee right"（有效）输入到我的查询中，我的查询应该 return ICD 代码 (IKR123) 与 DiagnosisName 列中的大部分单词相匹配。

以下是我当前的查询。（没有给出正确的输出）

SELECT diagnosisname
FROM
  (SELECT diagnosisname,
    UTL_MATCH.jaro_winkler_similarity('%operation Knee right%',diagnosisname)
  FROM icd_code
  ORDER BY UTL_MATCH.EDIT_DISTANCE_SIMILARITY('%operation Knee right%',diagnosisname) DESC
  )
WHERE ROWNUM<2;

此查询给我的输出为 "Left Knee Operation"，但我的期望是 "Right Knee Operation"。

Answer 1

请试试这个查询。这可能有助于解决您的问题。

SELECT diagnosisname 
   FROM (SELECT diagnosisname, UTL_MATCH.jaro_winkler_similarity('%operation Knee right%',diagnosisname) 
   FROM icd_code 
   WHERE UTL_MATCH.jaro_winkler_similarity('%operation Knee right%',diagnosisname) = 100
   ORDER BY UTL_MATCH.EDIT_DISTANCE_SIMILARITY('%operation Knee right%',diagnosisname) DESC) 
WHERE ROWNUM<2

Answer 2

关于您对 UTL_MATCH 的使用，有几点需要注意：

EDIT_DISTANCE_SIMILARITY : Returns 0 到 100 之间的整数，其中 0 表示完全不相似，100 表示完全匹配。
JARO_WINKLER_SIMILARITY : Returns 0 到 100 之间的整数，其中 0 表示完全不相似，100 表示完全匹配但尝试取考虑到可能的数据输入错误。

ORDER BY UTL_MATCH.EDIT_DISTANCE_SIMILARITY('%operation Knee right%',diagnosisname) DESC

这不会给你正确的结果。因为，您只考虑可能的相似性，而没有考虑数据输入错误。所以，你必须使用 JARO_WINKLER_SIMILARITY.

operation Knee right

您需要记住输入的 CASE 和要比较的列值。它们必须处于相似的情况下才能正确匹配。您在 LOWERCASE 中传递输入，但是，您的列值在 INITCAP 中。最好将列值和输入都转换为类似的情况。

让我们看看下面的演示来理解：

SQL> WITH DATA AS(
  2  SELECT 'Heart Operation' diagnosis_name, 'IH123' icd_code FROM dual UNION ALL
  3  SELECT 'Knee Operation' diagnosis_name, 'IK123' icd_code FROM dual UNION ALL
  4  SELECT 'Left Knee Operation' diagnosis_name, 'IKL123' icd_code FROM dual UNION ALL
  5  SELECT 'Right Knee Operation' diagnosis_name, 'IKR123' icd_code FROM dual UNION ALL
  6  SELECT 'Fever' diagnosis_name, 'IF123' icd_code FROM dual
  7  )
  8  SELECT t.*,
  9    utl_match.edit_distance_similarity(upper(diagnosis_name),upper('operation Knee right')) eds,
 10    UTL_MATCH.jaro_winkler_similarity (upper(diagnosis_name),upper('operation Knee right')) jws
 11  FROM DATA t
 12  ORDER BY jws DESC
 13  /

DIAGNOSIS_NAME       ICD_CO        EDS        JWS
-------------------- ------ ---------- ----------
Right Knee Operation IKR123         20         72
Knee Operation       IK123          20         70
Heart Operation      IH123          25         68
Left Knee Operation  IKL123         25         64
Fever                IF123          15         47

SQL>

因此，您会发现两者之间有何不同。 jaro_winkler_similarity 在识别 数据输入错误 和给出 最接近的匹配 [=38= 方面做得更好].在此基础上，只需选择第一行并按降序排列即可：

SQL> WITH DATA AS( 2 SELECT 'Heart Operation' diagnosis_name, 'IH123' icd_code FROM dual UNION ALL 3 SELECT 'Knee Operation' diagnosis_name, 'IK123' icd_code FROM dual UNION ALL 4 SELECT 'Left Knee Operation' diagnosis_name, 'IKL123' icd_code FROM dual UNION ALL 5 SELECT 'Right Knee Operation' diagnosis_name, 'IKR123' icd_code FROM dual UNION ALL 6 SELECT 'Fever' diagnosis_name, 'IF123' icd_code FROM dual 7 ) 8 SELECT diagnosis_name 9 FROM 10 (SELECT t.*, 11 utl_match.edit_distance_similarity(upper(diagnosis_name),upper('operation Knee right')) eds, 12 UTL_MATCH.jaro_winkler_similarity (upper(diagnosis_name),upper('operation Knee right')) jws 13 FROM DATA t 14 ORDER BY jws DESC 15 ) 16 WHERE rownum = 1 17 / DIAGNOSIS_NAME -------------------- Right Knee Operation SQL>

如何从 Oracle table 获取几乎匹配的字符串？

How to get almost matching string from Oracle table?

sql

oracle

string-matching

utl-match