Please start any new threads on our new
site at https://forums.sqlteam.com. We've got lots of great SQL Server
experts to answer whatever question you can come up with.
| Author |
Topic |
|
Blastrix
Posting Yak Master
208 Posts |
 Posted - 2003-05-28 : 23:17:49
|
| I'm currently needing to create a knowledge base for a help desk project I'm working on, and am looking for some suggestions. What I need to do is have a knowledge base, and when a user submits a support ticket. The body of the support request should be checked against the KB for potential articles that could answer their question before they submit the ticket.I have two initial thoughts. 1) Somehow use full-text indexing to look for possible matches, but I am not familiar with full-text indexing so I'm not quite sure how it would work.2) Use a more conventional method. Have a table of knowledge base articles, and then a have another table with article keywords. Take only known keywords out of the support request, and then run the found terms against the keywords table, inserting the results into a temp table or some such, and then ranking the articles to return by the number of entries in the temp table.Can anyone tell me which approach would be better, if a combination of the two would work, or if I'm completely off base?Thanks,Steve |
|
|
Merkin
Funky Drop Bear Fearing SQL Dude!
4970 Posts |
Posted - 2003-05-29 : 00:26:55
|
| Maybe some combination of the two.Do a full text catalogue of your articles. Then mainatin a table of "Interesting" words. When someone enterers a request, find the Interesting words inside that request and pass them to the text search.Might be worth a shot.Damian |
 |
|
|
Blastrix
Posting Yak Master
208 Posts |
Posted - 2003-05-29 : 11:08:06
|
| I was kind of thinking about that option, but there might be one small problem. If I have a list of keywords that I try to pick out of a request, I miss any type of inflectional changes as I would get with a full-text query. Granted I could still check inflectional differences between the found keywords, and those words in the articles, but if one of my keywords was "took", and the request contained "taken" I would completely miss it. Any ideas to get around that hurdle?Steve |
|
|
|
|
|
|
|
|
Creating Knowledge Base(Search Options)