thanks to David we are going a bit further with the tests.
I am writing about the formula used in the 'date factor' weight. Can
you please point me to some reference? I find this formula a bit 'out of
tune', as old documents (currently earlier than 2000 - appx.ly) easily
disappear from the results as the factor gets a negative score
(amplified by the configuration value of the factor).
As I wrote in the SF task, the actual formula is something like this:
date_score = date_factor * ((document_date * 10 / now) - 9)
From: Lachlan Andrew <lha@us...> - 2004-03-14 06:50:56
A candidate would be
date_score =3D date_factor*100 / (1 + (now-document_date)/356*24*60*60)
That says that something a year old is half as relevant as something=20
hot off the press, and nothing gets zero/negative.
On Thu, 11 Mar 2004 22:15, Gabriele Bartolini wrote:
> I am writing about the formula used in the 'date factor' weight.
> date_score =3D date_factor * ((document_date * 10 / now) - 9)
ht://Dig developer DownUnder (http://www.htdig.org)