Question

在PL / SQL中进行单词匹配搜索的最佳方法是什么？

E.g。字符串“ BROUGHTONS OF CHELTENHAM LIMITED ”

“BROUGHTONS LIMITED”是匹配

“OF LIMITED”是匹配

“CHELTENHAM BROUGHTONS”是匹配

“BROUG”是不匹配

Answer 1

这是一种相当粗略的方法，但应该按照你的要求行事。正如Xophmeister所指出的那样，你可能需要对每个字符串进行标记，然后搜索标记（因为你想要不按顺序匹配，做一个简单的“像％tokenA％tokenB％tokenC％”将无效）。

此外，这甚至没有涉及语音，soundex等所有问题。但是，再次，不是你问的问题。这也不会触及性能或扩展问题，并且可能只对少量数据可接受。

所以，首先我们需要一个分割函数：

create or replace 
function fn_split(i_string in varchar2, i_delimiter in varchar2 default ',', b_dedup_tokens in number default 0)
return sys.dbms_debug_vc2coll
as
  l_tab sys.dbms_debug_vc2coll;
begin
  select regexp_substr(i_string,'[^' || i_delimiter || ']+', 1, level)
  bulk collect into l_tab
  from dual
  connect by regexp_substr(i_string, '[^' || i_delimiter || ']+', 1, level) is not null
  order by level;

  if (b_dedup_tokens > 0) then
    return l_tab multiset union distinct l_tab;
  end if;
  return l_tab;
end;

现在我们可以用它来检查特定标记的字符串。在这里，我正在从一组数据样本中搜索3个令牌（John Q Public）

with test_data as (
  select 1 as id, 'John Q Public' as full_name from dual
  union
  select 2 as id, 'John John Smith' as full_name from dual
  union
  select 3 as id,'Sally Smith' from dual
  union
  select 4 as id, 'Mr John B B Q Public' from dual
  union
  select 5 as id, 'A Public John' from dual
)
select d.id, d.full_name, count(1) as hits
from test_data d, table(fn_split(full_name, ' ', 1))
-- should have at least 1 of these tokens
where column_value in ('John', 'Q', 'Public')
group by d.id, d.full_name
-- can also restrict results to those with at least x token hits
having count(1) >= 2
-- most hits at top of results
order by count(1) desc, id asc

输出：

"ID"    "FULL_NAME" "HITS"
1   "John Q Public" 3
4   "Mr John B B Q Public"  3
5   "A Public John" 2

您还可以添加“upper”以区分大小写，等等。

Answer 2

使用Oracle Text索引。这将允许您发出强大的CONTAINS查询。

http://docs.oracle.com/cd/B28359_01/text.111/b28303/quicktour.htm

使用PL / SQL进行单词匹配搜索

2 个答案: