Changing the Language Noise Words

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Darren Jensen - 17 Nov 2004 11:45 GMT

Hi,

I have searched high and low for an answer to this question. Please
tell me that it is a simple process....!

I have created a website in English using SQL 2000 as the database.
Now I have just launched the same website in French using exactly the
same schema (and same server) as the english website. Both use
full-text search. BUT...I notice that the noise word file being used
for both is English so, naturally, I need to change my French language
site to French. How can I do this?

I have already tried changing the database and field collation in the
French database and re-creating and re-populating the full-text index
but the English noise word file is still being used. Please, somebody
help!!

Reply to this Message

Gerald Baeck - 17 Nov 2004 15:43 GMT

Darren,

You can switch to another language via dropping the existing FT Catalog and
re-creating it via the FT Indexing Wizard and when you pick the column (Text)
to be FT Indexed, click on the drop down box marked "Language for Word
Breaker" and select your language and then run a full population.
(Thanks to John for telling that trick!-)

Regards, Gerald.

> Hi,
>
[quoted text clipped - 12 lines]
> but the English noise word file is still being used. Please, somebody
> help!!

Reply to this Message

John Kane - 17 Nov 2004 16:29 GMT

Darren,
Garald is correct. Just some additional info for future reference... You can
find out which noise.* is linked to what language via the Schema.txt file
under \FTDATA\SQLServer\Config, for example for noise.FRA is the noise word
file for French:

<stoplist
language="French_French"
file="noise.fra"
primarylanguage=12
sublanguage=1>

Regards,
John

> Darren,
>
[quoted text clipped - 22 lines]
> > but the English noise word file is still being used. Please, somebody
> > help!!

Reply to this Message

Darren Jensen - 18 Nov 2004 03:19 GMT

Thanks, that worked a treat and very easy to do too!

> Darren,
>
[quoted text clipped - 5 lines]
>
> Regards, Gerald.

Reply to this Message

Darren Jensen - 23 Nov 2004 04:26 GMT

I have one follow up question on the subject of fulltext indexing with
French and that is how can I get the search to ignore accents? For
example if I search 'hopital' I would like the results to show all
places where it finds 'hôpital' or 'hopital'. Is this possible?

Thanks.

> Thanks, that worked a treat and very easy to do too!
>
[quoted text clipped - 7 lines]
> >
> > Regards, Gerald.

Reply to this Message

John Kane - 23 Nov 2004 05:27 GMT

Darren,
Unfortunately, the issue of accent sensitivity vs. accent insensitivity FTS
is more difficult to answer. This is a well known bug that has been around
for a long time (since SQL Server 7.0) and is truly only going to be fixed
in SQL Server 2005 <sigh>. As there are only workarounds for this bug in SQL
Server 2000 that require the use of duplicate data where you store the
non-accented search words, such as 'hopital'. You would then FT Index and FT
Search this column, while returning the accented search data (such as
'h?pital') from another column back to the user. You will need to develop
accent removal procedures as well as use triggers (insert & update) to
maintain the currency of the two columns....

Regards,
John

> I have one follow up question on the subject of fulltext indexing with
> French and that is how can I get the search to ignore accents? For
[quoted text clipped - 14 lines]
> > >
> > > Regards, Gerald.

Reply to this Message

Gerald Baeck - 23 Nov 2004 07:28 GMT

Darren,

my workaround for this problem is to replace all special chars before
filling the index. I know it could take a while, but till Yukon is released
its the only way i think.

UPDATE [table] SET [field] = REPLACE([field], '?', 'o')

Regards, Gerald.

>I have one follow up question on the subject of fulltext indexing with
> French and that is how can I get the search to ignore accents? For
[quoted text clipped - 16 lines]
>> >
>> > Regards, Gerald.

Reply to this Message