Sorry for the multiple message, Markus and Lewis. I did read the wiki entry on NewScoring. It is described as a standalone process AFTER crawling. So what does the scoring-link plugin do? Is it equivalent to NewScore or not?
On Sun, Jun 16, 2013 at 12:21 PM, Joe Zhang <[email protected]> wrote: > and scoring-link != LinkRank? > > > On Sun, Jun 16, 2013 at 12:20 PM, Joe Zhang <[email protected]> wrote: > >> So this is a stand-alone process after the crawling. >> >> >> On Sun, Jun 16, 2013 at 12:17 PM, Markus Jelsma < >> [email protected]> wrote: >> >>> Hi Joe, >>> >>> You don't need a scoring filter for Linkrank. Just follow the wiki and >>> run the webgraph tool on your segments. Then you can run the linkrank tool >>> on the webgraph you just created from your segments. Finally use the >>> scoreupdater tool to write the scores back to your crawldb. >>> >>> Cheers >>> >>> https://wiki.apache.org/nutch/NewScoring >>> >>> >>> -----Original message----- >>> > From:Joe Zhang <[email protected]> >>> > Sent: Sun 16-Jun-2013 21:14 >>> > To: user <[email protected]> >>> > Subject: Re: Nutch scoring question again >>> > >>> > Is scoring-link preferred over scoring-opic? I saw some disucssion of >>> > deficiencies of opic. >>> > >>> > >>> > On Sun, Jun 16, 2013 at 12:12 PM, Lewis John Mcgibbney < >>> > [email protected]> wrote: >>> > >>> > > Yes Joe this is correct. >>> > > >>> > > >>> > > On Sun, Jun 16, 2013 at 12:03 PM, Joe Zhang <[email protected]> >>> wrote: >>> > > >>> > > > Thanks. >>> > > > >>> > > > with regards to (2), is this score the "boost" we see in solr >>> index? >>> > > > >>> > > > >>> > > > On Sun, Jun 16, 2013 at 10:38 AM, Ahme Emre Aladağ >>> > > > <[email protected]>wrote: >>> > > > >>> > > > > Note: I'm a newbie. >>> > > > > >>> > > > > As far as I know, new scoring and scoring-link corresponds to >>> LinkRank. >>> > > > > It's implemented in the scoring.webgraph package. The code in the >>> > > > > scoring-link might be linking the scoring plugin system to the >>> LinkRank >>> > > > > class in webgraph. >>> > > > > >>> > > > > 1) Yes it works for sorting the pages. The topN most >>> important-seeming >>> > > > > pages are fetched in the next cycles according to this scoring. >>> > > > > 2) Relevance in retrieval is affected due to (1). It calculates >>> the >>> > > > scores >>> > > > > and gives them to Solr. Solr will rank the search results >>> according to >>> > > > > these scores and some other external custom scores. >>> > > > > >>> > > > > >>> > > > > ----- Orijinal Mesaj ----- >>> > > > > Kimden: "Joe Zhang" <[email protected]> >>> > > > > Kime: "user" <[email protected]> >>> > > > > Gönderilenler: 15 Haziran Cumartesi 2013 23:41:33 >>> > > > > Konu: Nutch scoring question again >>> > > > > >>> > > > > The plugins directory only contains two scoring plugs: >>> scoring-link and >>> > > > > scoring-opic. What about the newscoring, linkrank, etc.? Where >>> are they >>> > > > > available? >>> > > > > >>> > > > > Again, I'm confused about the nature/purpose of such scoring: >>> > > > > >>> > > > > 1. Does it work as a sorting function for the frontier of the >>> crawling? >>> > > > --> >>> > > > > this seems reasonable. >>> > > > > 2. Or does it affect relevance in retrieval? If so, why is it >>> handled >>> > > in >>> > > > > the crawler, but not solr? >>> > > > > >>> > > > > I'd greatly appreciate any enlightment. >>> > > > > >>> > > > >>> > > >>> > > >>> > > >>> > > -- >>> > > *Lewis* >>> > > >>> > >>> >> >> >
