I share your sentiment about Verity's capability though I appreciate its speed. And that's the very reason why I tried something else.
The semantic search technique that I'm using now, imho, provides better search results, record rendering even with cf4.5 against a ms sql server 2000 database isn't bad. With cf8 and full-text search w/ ms sql server 2000/2005 record rendering speed SUBSTANTIALLY improves (I could test to see if cf8 helps in this regard but of little interest to me for now). I find even optimized t-sql can't beat freetext indexing. Not preaching my ware, just an opinion. Having said the above, apple, orange, banana, you name it, to everyone's opinion and liking... Happy searching, everyone. > In Verity, you CAN search within the "customX" fields as well as all > other fields (category, author, etc...). > > Verity is, however, mostly useless unless you are trying to find exact > matches. The ranking engine in verity is the equivalent of the > "FindNoCase" function in CF. It will return a T/F. Thats it. There is > no ranking in Verity based on the NUMBER of times your search term was > matched (or anything else useful such as "<h1>" or "<title>"). > > Verity automatically provides stemming (horse -> horses and vice > versa) which is nice BUT it ranks matches the same (horse is worth the > same as horses). It supports thesaurus (fast -> quick and vice versa), > but the same ranking problem exists and MS SQL supports theasurus as > well (without the ranking problem). > > In all, I would recommend against using Verity under any/all > circumstance if there is even the slightest alternative. We just spent > MANY hours building a robust search for a client based on the VERITY > engine and we deeply regret it. > > Everything could have been done in SQL (SQL 2005 also has a VASTLY > superior ranking engine to Verity) and then we could have simplified > JOINs and other filters. Add the utility of the JOINs to a 20x (at > least) performance INCREASE in MS SQL and you have a clear winner. > > If your content is specifically web sites, I'd also recommend checking > out the google search API and/or google Co-op. > > Avoid Verity if you have ANY alternative options and you don't hate > yourself. > > > > > Hi, > > > > (sorry if this post appears twice, having problems with the > submission > > page and our firewall) > > > > Iâm looking for other peoples opinion on searching via a database vs. > > > Verity searches. > > > > Iâm building new search pages for our course information system, > with > > the end result being website and course information results being > > combined (if required) or filtering the results based on things like, > > > course information only, campus location, course type e.g. > mainstream, > > application only courses etc. > > > > Currently the course information search pages are different to the > > site search. Course Information searches an MSSQL database whereas > > > the site search uses a couple of verity collections (via FarCry). > > > > The Verity engine is a little more limited than what I originally > > thought, for example: you can only search the contents of the body, > > > not the custom1, 2, 3 and 4 fields or the title field. I did have a > > > look at using the new Category feature, but aside from it still > being > > too limited (by the amount of fields I possibly need to filter by) > > itâs considerably slower to search than when not using categories. > > > Using a small sample of 1250 records, searching without categories > > took an average of 150ms, when restricting the same search using > > categories it took on average 3500ms. > > So where Iâm up to now is either > > > > 1) Not using Verity for the course information and going back to > > searching against the database. I could split the site search > result > > page to display two separated results, one based on the database > > (course info) and the other on the verity collection (website). You > > > then would select which set of results to continue on with for page > 2. > > > > > > 2) Building x number of verity collections for each type of course > > information search (application courses, mainstream etc) and combine > > > these when searching. The would mean that the body of all these > > collections would have identical information, which just feels wrong. > > > > > 3) Attach the different types of filtering options (as keywords) to > > > the end of the body section with some type of marker to let me know > > > from that point on is filtering keywords and to strip them out > before > > displaying the text on the results page. I need to give this one a > > > bit more thought, but this doesnât feel very future proof. > > > > Any other suggestions? > > > > Thanks for your time. > > > > Regards, > > > > Mark Picker > > Internet Developer - External Business Systems > > http://www.wit.tafensw.edu. au/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Download the latest ColdFusion 8 utilities including Report Builder, plug-ins for Eclipse and Dreamweaver updates. http;//www.adobe.com/cfusion/entitlement/index.cfm?e=labs%5adobecf8%5Fbeta Archive: http://www.houseoffusion.com/groups/CF-Talk/message.cfm/messageid:294483 Subscription: http://www.houseoffusion.com/groups/CF-Talk/subscribe.cfm Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.4
