"brasil" will return companies with this word in any part of its name. this search (randomic in 4 different levels) is only for promoted records (10000 records to be searched at all). free records (10 milion) are the fifth level and will respect the common search mode.
Um Abraço, Paula C. Laun : Dataprisma pa...@dataprisma.com.br (47) 3035.1868 www.dataprisma.com.br ----- Original Message ----- From: "Walter Underwood" <wun...@wunderwood.org> To: <solr-user@lucene.apache.org> Sent: Monday, December 20, 2010 2:31 PM Subject: Re: about groups of random results + alphabetical result The problem happens with any common word, not just short words. What happens with "Brasil"? If this was a good way to do search, Solr would already implement it. It is not that hard to build. But it is not a good way to do search. I have been working on search for almost 15 years, and I hear this idea every year or two. Don't do it. Use the QueryElevationComponent for step 1, boots in DisMax for steps 2-4, and don't do step 5. People will never scroll down that far, besides, phonetic search will match a lot of the documents. wunder On Dec 20, 2010, at 8:09 AM, Paula C. Laun : Dataprisma wrote: > thank you for your help... this search will be published in Portuguese, > and > in this language we can clean up the sentence from words shorter than 3 > characters. > > Paula C. Laun : Dataprisma > pa...@dataprisma.com.br > (47) 3035.1868 > www.dataprisma.com.br > ----- Original Message ----- > From: "Walter Underwood" <wun...@wunderwood.org> > To: <solr-user@lucene.apache.org> > Sent: Monday, December 20, 2010 2:02 PM > Subject: Re: about groups of random results + alphabetical result > > > You probably do not want this ranking, because any query with a common > word, > like "the", will match most of the corpus in step two. > > Instead, use Solr to weight better quality matches more heavily, maybe 4X > for exact matches, 2X for stemmed matches, and 1X for phonetic matches. > > wunder > > On Dec 20, 2010, at 4:01 AM, Paula C. Laun : Dataprisma wrote: > >> hi. i'm looking for a technology who could have high performance in >> searching a high amount of data (nearly 10 milion lines in a >> convencional >> database like sql server) and i think PHP running under apache solr is a >> good choice. i have only a doubt about its possibilities. >> >> i need to show in first place: promoted records who have all the terms >> searched by the user (ordered randomly). >> in second place: i need to show promoted records who have any term >> searched >> by the user (ordered randomly). >> in third place: i need the promoted records found by the stemming search >> (ordered randomly). >> in fourth place: i need the promoted records found by the phonetic search >> (randomly). >> in fifth place: the free records ordered alphabeticly. >> >> these results need to be paginated. >> >> is it possible to do that in the same task? >> >> Thanks, >> >> Paula > > > > > -- Walter Underwood Venture ASM, Troop 14, Palo Alto
