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

[<kau...@cs...>]

On 11/7/2005, "stack" <st...@ar...> wrote:
> >Many news sites have a structure which may distort ranking of search
> >results. On these sites each page focuses on one current news item but
> >it also has loads of tiny links to other news, like 'other top
> >stories' , 'news of previous day'  or 'stories of this week'.
> Just to be clear, in the above, you mean that the outlink anchor text
> says 'other top stories' and 'stories of this week'?

Hi,
I tried to describe pages and links like

------- Page start -------
<h1>English cricket results 9.11.2005</h1>
........ lots of text .....

<h2>Other top stories now</h2>

<a href=3D"http://www.tekstia.com/news/top/earthquake+timbuktu/107994">
Earthquake in Timbuktu second time this year</a>
<a href=3D"http://www.tekstia.com/news/top/tokyo+stocks+explode/107996">
Tokyo stocks explode</a>

10 further links to different subjects ..
---- Page end -------------

Above the links are <a href=3Durl>text</a> , and both text and url contain
words which actually don't belong to the body text of the cricket news
article.

> >In an indexed archive, you could search for 'earth quake' and find
> >close  to top search results a page with headline 'cricket results'.
> >Reason being that the cricket page has one link to earth quake news of
> >previous day.
> >
> >This feature is noticeable when there are few pages with real earth quake
> >reports  but lots of other pages having links to them.
> >
> >
> >
> Yes.  Makes sense.
>
> You can set the boost on inlink anchor text -- see the just-added FAQ --
> but looks like you want to be able to set separately the boost on a
> documents' outlink anchor text.  Looking at the Nutch html parser code,
> currently the outlink anchor text just gets added to the StringBuffer
> accumulating all document parsed 'text'; the outlink anchor text is not
> distingushed in any way from the general text of the document.  There is
> currently no means of making its boost be different from that of the
> general document 'text'.
>=20
> Should we add such a feature Kaisa?