Re: [Archive-access-discuss] Ranking search results ?

[stack <st...@ar...>]

kau...@cs... wrote:

>Is it possible to rank/sort search results by relevance? Show first
>results
>where search term is in html title, or appears several times in text.
>(versus those where search term appears once, late in text, or
>in a link name).
>  
>
It should be doing this for you Kaisa.  In general, are you not seeing 
the most significant links showing first in results?

I just added a little FAQ on ranking with some notes on how nutch is 
doing it.  I'll repeat the note here:

By default, at query time, the following fields are boosted as follows:

query.url.boost, 4.0f
query.anchor.boost, 2.0f
query.title.boost, 1.5f
query.host.boost, 2.0f
query.phrase.boost, 1.0f

 From the above, terms found in an URL are scored high with anchor text 
next, then title.

You can change the above boosts by editing your nutch-site.xml but in 
general, the defaults seem to work well for most collections.

Anchor text can make a large contribution to a document ranking score. 
You can see the anchor text for a page by browsing to the 'explain' then 
editing the URL to put in place 'anchors.jsp' instead of 'explain.jsp'.

>Does nutchwax index link names within html files? If there's a link
>http://www.something.net/storm.gif withing html , could I search for
>'storm'
>and get this image into result list?
>  
>
This is an interesting question Kaisa.  I just took a look.  It doesn't 
look like it (See below for how I figured this).  Do you need this feature?

Here's how I took a look see at what was in the a particular nutch segment:

%  ./bin/nutch segread -fix -nocontent -dump 
nutch-data/segments/debord2005-11-01-155531/

This dumps out what nutch has per resource.  It will list the text it 
parsed from the document, list of outlinks found in the document, the 
page hash, etc.

I compared what was in nutch to what was in the indexed ARC (I zcat'd 
the ARC).

Yours,
St.Ack

>*Kaisa
>
>
>-------------------------------------------------------
>This SF.Net email is sponsored by the JBoss Inc.
>Get Certified Today * Register for a JBoss Training Course
>Free Certification Exam for All Training Attendees Through End of 2005
>Visit http://www.jboss.com/services/certification for more information
>_______________________________________________
>Archive-access-discuss mailing list
>Arc...@li...
>https://lists.sourceforge.net/lists/listinfo/archive-access-discuss
>  
>