Full Text Search...Not Impressed

[neal007]

Hello,

After upgrading several forums, some quite large, to 3.5, we switched all to use the Full Text Search (We're using PHP 5.0.5, MySQL 4.1.14 on Windows Server 2003/IIS 6).

Visitors were complaining about search results and I ran a few tests, such as searching my forum (www.nc-software.com/forums) on something as simple as "121" (without quotes). I tried wildcards such as "*121*" and still did not get any results returned. I switched back to vBulletin search and tried "121" and did find the threads I was looking for.

Two large forums would certainly benefit (http://forums.flightinfo.com) (other private) but the Full Text Search has not proven a viable search system, unless anyone can recommend a method to improve it.

We switched back (and are currently running the update counters/rebuild search index) to vBulletin searching until further notice.

FYI

[Zonex]

i think i remember one of the developers saying thats an sql setting for minimum characters or something
nothing vb can change, only your host or yourself if u are your own host

[Zachery]

The MySQL mininum size word by default I believe is 4, and after you switched you did run the nessary queries to index the threads?

[neal007]

The MySQL mininum size word by default I believe is 4, and after you switched you did run the nessary queries to index the threads?

I'm looking in MySQL Admin and the MyISAM properties and see the minimum word length is 4, but the area is grayed out, so I can't change it.

When I switched to Full Text Search I did the defaults from vBulletin, i.e. ran the queries (I think it said there were two).

So now, figuring out how to drop the word length to 3....

[neal007]

Found this...shows how to reduce word length, will try:

http://dev.mysql.com/doc/mysql/en/fu...ne-tuning.html

[Zachery]

You will have to drop the fulltext indcies and rebuild them.

[neal007]

You will have to drop the fulltext indcies and rebuild them.

How do you do this? It would be nice (suggestion) if this was in the SQL Query section of the Maintenance area or update counters section.

[Zonex]

Right under main options where you switch search types. Kind of an odd place for it.

[neal007]

Right under main options where you switch search types. Kind of an odd place for it.

Yes, I know that, but that's an ADD SQL statement, I don't know that it's doing a drop and rebuild/reindex. I just want to be sure I'm doing the right thing.

By the way, I have the "Quote message in reply" checked on the bottom left of the quick reply...and I notice as I type it's toggling checked and unchecked...odd!

[Zonex]

I think you need to switch back and forth from search types. That's the only place that I see where it rebuilds indices...when you switch it presents you with a build/drop option.

[neal007]

I think you need to switch back and forth from search types. That's the only place that I see where it rebuilds indices...when you switch it presents you with a build/drop option.

Thanks, that's what I've been doing. Okay, will see how things go now with word length at 3.

[Freddie Bingham]

You may very well lose all performance benefits of Fulltext by dropping the minimum word length from 4 to 3.

[neal007]

You may very well lose all performance benefits of Fulltext by dropping the minimum word length from 4 to 3.

Well...a lot of aviation and government agencies use 3 letter acronyms, it's like a language of their own. Not being able to seach on 3 letters would make the search system non-effective. Some boards are big, I understand one benefit of the FT search is not having horrendously sized search index tables. It also takes a very long time to re-index these sites (6-8 hours) so, after switching to FT search, it's been difficult getting reindexed without some issue such as a page not found error coming up, time out of some sort, etc.

I'll see how the word length of 3 works out. I already noticed I'm not able to find anything on searching for "far" at http://forums.flightinfo.com where there are a lot of posts with "far" (or FAR) in it, as the FAR is the acronym for Federal Aviation Regulation and a topic of discussion quite often.

[Freddie Bingham]

There may also be three letter words in MySQL's stop list, which would have to also be removed. At last read that requires editing a file and recompiling MySQL.

[neal007]

There may also be three letter words in MySQL's stop list, which would have to also be removed. At last read that requires editing a file and recompiling MySQL.

I edited the My.INI and restarted the MySQL service (Windows 2003) and when running MySQLAdmin it shows 3 now, and not disabled/grayed out like before, so it's detecting the new setting properly. I tested on another forum and it is working (searching on 121 at http://www.nc-software.com/forums). I think I need to remove and redo the FT search on Flightinfo, but every time I do this is shuts down all forums (database error) as it is too intensive of an operation I believe (and it's a powerful server configuration, dual xeon for web (forum) and dual PIII for separate DB server, 4 Gigs of RAM each).