Home » Server Options » Text & interMedia » Context IndexType & can I use contains in searching for Arabic Values?
Context IndexType & can I use contains in searching for Arabic Values? [message #387266] Wed, 18 February 2009 10:48 Go to next message
ledo
Messages: 7
Registered: February 2009
Junior Member
I have a problem in using IndexType (CONTEXT) in searching for Arabic value by (contains).

when I used it to get matched values for an English text, it gave a good results and the nearest matched values not the exact ones only (alternatives by fuzzy matching), but using the same select statement but this time for getting the matched records of an Arabic text, it was just returned records (results) if and only if I passed the exact values, no fuzzy or alternatives...

SELECT score(1),
AR_COL
FROM places
WHERE CONTAINS (AR_COL, '
<query>
<textquery>
<progression>
<seq>عسى</seq>
</progression>
</textquery>
</query>'
, 1) > 0 ORDER BY score(1) DESC;
)

So values returned are those which are the same as 'عسى'(exact matching), and if I changed/wrote any other text near to this one and used '?' to set the fuzzy matching, the results are (No rows returned)

so how can I use the context index in matching Arabic values, and to be more accurate, how can I get fuzzy match of each word?

Thanks in advance,
Re: Context IndexType & can I use contains in searching for Arabic Values? [message #387340 is a reply to message #387266] Wed, 18 February 2009 21:53 Go to previous messageGo to next message
Barbara Boehmer
Messages: 9084
Registered: November 2002
Location: California, USA
Senior Member
Oracle Text does not support fuzzy definitions for Arabic. The following is an excerpt from the section of the online documentation in the link below.

"Oracle Text supports fuzzy definitions for English, French, German, Italian, Dutch, Spanish, Portuguese, Japanese, OCR, and auto-language detection."

http://download.oracle.com/docs/cd/B28359_01/text.111/b28304/cqoper.htm#CCREF0307

Re: Context IndexType & can I use contains in searching for Arabic Values? [message #387956 is a reply to message #387340] Sun, 22 February 2009 11:25 Go to previous messageGo to next message
ledo
Messages: 7
Registered: February 2009
Junior Member
Great,

First thanks for your reply.

Second is there no way to make a "fuzzy matching" with my context index on Arabic words?, I mean what can I do to get the alternatives/matched words in a specific column (Arabic values) for a search text?

I read something about the WORLD_LEXER as a preference and supporting Arabic and the searching didn't come with any useful decision, can you please help in my case?

Also Basic_WORDLIST for prefix and substring indexing, if I used this one, it will give better accurate results than using LIKE operator?
Re: Context IndexType & can I use contains in searching for Arabic Values? [message #387973 is a reply to message #387956] Sun, 22 February 2009 13:43 Go to previous message
Barbara Boehmer
Messages: 9084
Registered: November 2002
Location: California, USA
Senior Member
Quote:

Is there no way to make a "fuzzy matching" with my context index on Arabic words?, I mean what can I do to get the alternatives/matched words in a specific column (Arabic values) for a search text?



The only thing I can think of is to create your own thesaurus using ctx_thes.create_relation and add the matching words in pairs using ctx_thes.create_relation, making them synonyms, so that you can use ctx_thes.syn on your search string to obtain all of the possibilities.

Quote:

I read something about the WORLD_LEXER as a preference and supporting Arabic and the searching didn't come with any useful decision, can you please help in my case?



The world_lexer supports Arabic, as stated in the link below, but not fuzzy.

http://download.oracle.com/docs/cd/B28359_01/text.111/b28304/amultlng.htm#CEGBAFHC

Quote:

Also Basic_WORDLIST for prefix and substring indexing, if I used this one, it will give better accurate results than using LIKE operator?



You will get the same results either way, but if you are using a wildcard prefix you will likely get faster results with substring index.

Previous Topic: What is Score(1) in Oracle Text?
Next Topic: CATSEARCH with Fuzzy Matching...?
Goto Forum:
  


Current Time: Tue Apr 16 09:02:20 CDT 2024