Re: Add link popularity to indexer
Peter Karman <[email protected]>
| Newsgroups | gmane.comp.web.swish-e |
|---|---|
| Message-ID | <[email protected]> |
this is a oft-requested feature. search the mail archive for other threads. one way to approach this is to add a property via a filter/-S prog combination and then sort first by that property, secondly by rank. Cas Tuyn scribbled on 7/27/06 7:04 AM: > Hi, > > We have implemented Swish-e via the spider method, and notice the > results are not ranked according to the most useful first. We already > gave extra points to the title, description and H1,H2,H3. > > Every project in our company has a welcome page, with links to the > team members, background literature, project documentation, output > examples and deliverables. Ideally if people search for the project > name, they should find the project welcome page, but instead they find > the detailed pages below the project, because they mention the project > name more often, and have the project name in their title and > description too. > > I already went as far to remove the descriptions from the underlying > pages, so that the welcome page would have a 8 point headstart, but > that's not enough to raise the welcome page to the #1 position. > > Is there a way to import traffic server logs into the index results, > so that whatever is visited most, gets extra points? > > We certainly do not want to emulate Google where popularity is based > on the amount of links into a page, because our intranet is a means, > not a goal by itself. Page owners must write down their knowledge, and > not worry about raising their or other people's pages in the search > engine. > > One way could be that the distance from the spider home page which is > the indexing start point adds a multiplier to all results: > - indexer start = times 6 > - level 1 = times 5 > - level 2 = times 4 > - level 3 = times 3 > etc. > > Looking forward to your replies and suggestions... > > Cas > -- Peter Karman . http://peknet.com/ . [email protected]