Re: Relevancy Rating
This WebDNA talk-list message is from 1998
It keeps the original formatting.
numero = 21787
interpreted = N
texte = >>I have written a sherlock plugin and want to utilize the relevance ranking >>that >>sherlock can identify. How can I create a search result that has a numeric >>relevancy given with each founditem? Sherlock looks for a 100 for 100% >>relevancy on down to 0 for least relevant. > > Unfortunately WebCatalog does not expose its relevancy information inside > [founditems], but you can fake one by assuming that the topmost items are > the most relevant, and using a little [math] with the [index] and > [numfound] to create a smooth gradient from 100% down to 0% (I guess 50% or > so would make more sense). > > Sherlock is very nice, and very easy to implement -- congrats to Apple! > > Technical Support | ==== eCommerce and Beyond ==== I actually use the # of search terms to calculate a simple relevancy rating for our Sherlock plugin.predator = 1 word = 25% predator aliens = 2 words = 50% ... 4 or greater words = 100%It's not perfect (all items in a single search instance have the same relevance), but it's at least a rough measurement of overall relevance. Matching 4 words is probably more relevant than matching only 1 word in the grand scheme of things...Any chance a future version of Webcat will expose relevance inside [founditems]?-Dale ________________________ ______________________________ | Dale LaFountain | mailto:dalel@dhorse.com | | IT Director, WebMaster | http://www.dhorse.com/ | | Dark Horse Comics, Inc.|Phone: 503-652-8815 x324 | \________________________|______________________________/
Associated Messages, from the most recent to the oldest:
>>I have written a sherlock plugin and want to utilize the relevance ranking >>that >>sherlock can identify. How can I create a search result that has a numeric >>relevancy given with each founditem? Sherlock looks for a 100 for 100% >>relevancy on down to 0 for least relevant. > > Unfortunately WebCatalog does not expose its relevancy information inside >
[founditems], but you can fake one by assuming that the topmost items are > the most relevant, and using a little
[math] with the [index] and > [numfound] to create a smooth gradient from 100% down to 0% (I guess 50% or > so would make more sense). > > Sherlock is very nice, and very easy to implement -- congrats to Apple! > > Technical Support | ==== eCommerce and Beyond ==== I actually use the # of search terms to calculate a simple relevancy rating for our Sherlock plugin.predator = 1 word = 25% predator aliens = 2 words = 50% ... 4 or greater words = 100%It's not perfect (all items in a single search instance have the same relevance), but it's at least a rough measurement of overall relevance. Matching 4 words is probably more relevant than matching only 1 word in the grand scheme of things...Any chance a future version of Webcat will expose relevance inside
[founditems]?-Dale ________________________ ______________________________ | Dale LaFountain | mailto:dalel@dhorse.com | | IT Director, WebMaster | http://www.dhorse.com/ | | Dark Horse Comics, Inc.|Phone: 503-652-8815 x324 | \________________________|______________________________/
Dale LaFountain
DOWNLOAD WEBDNA NOW!
Top Articles:
Talk List
The WebDNA community talk-list is the best place to get some help: several hundred extremely proficient programmers with an excellent knowledge of WebDNA and an excellent spirit will deliver all the tips and tricks you can imagine...
Related Readings:
quotes and truncating? (1997)
WebCat2 beta 11 - new prefs ... (1997)
M$loth messes with our sites (again) 2004/02/03 (2004)
Store Builder, technical infos (1998)
WebCat2 - [format thousands] (1997)
Max Record length restated as maybe bug (1997)
Req: Array Tags Explained, simple and clear (2004)
month or year from date field in lineitems? (1998)
Up and running ... at last !! (1997)
Security (2006)
Thanks ! (1997)
available times? (2003)
RE: Adding headers to email (1997)
multiple credit card merchant accounts/processors (1997)
Re:No Data (1997)
Multiple prices (1997)
synching databases on multiple servers (1997)
Sort Issue? (2000)
PCS Frames (1997)
suffix mapping for NT? (1997)