| |
In lugnet.admin.general, Dan Boger writes:
> In lugnet.admin.general, Todd Lehman writes:
> > http://www.lugnet.com/spotlight.cgi
>
> Great work :) just curious, how is it sorted? It seems to be sorted by
> elapsed time...
See text at top of page there.
> shouldn't it sort by rating, perhaps with a time formula to lower the
> rating?
It would quite easy to offer a sorting choice --
a) Newest first ("weblog" style)
b) Most recommended first ("weekly top 40" style)
c) Fuzzy combination of both ("today's top stories" style)
--Todd
|
|
| |
Todd Lehman wrote:
> > > http://www.lugnet.com/spotlight.cgi
> >
> > Great work :) just curious, how is it sorted? It seems to be sorted by
> > elapsed time...
>
> See text at top of page there.
doh, yup... it's right there.
> > shouldn't it sort by rating, perhaps with a time formula to lower the
> > rating?
>
> It would quite easy to offer a sorting choice --
>
> a) Newest first ("weblog" style)
> b) Most recommended first ("weekly top 40" style)
> c) Fuzzy combination of both ("today's top stories" style)
mmm... fuzzy :) by day then by rating? nod... even put a separator
between days, somewhat like the current (Static) spotlight...
:)
Dan
|
|
| |
In lugnet.admin.general, Dan Boger writes:
> > > shouldn't it sort by rating, perhaps with a time formula to lower the
> > > rating?
> >
> > It would quite easy to offer a sorting choice --
> >
> > a) Newest first ("weblog" style)
> > b) Most recommended first ("weekly top 40" style)
> > c) Fuzzy combination of both ("today's top stories" style)
>
> mmm... fuzzy :) by day then by rating? nod... even put a separator
> between days, somewhat like the current (Static) spotlight...
Hmm...Maybe, but I meant fuzzy in the fuzzy-logic sense -- take the internal
score (0-100) and convert it to a real number in the unit interval [0,1], then
take the age of the message relative to the specified time period and make
that also a number in the unit interval [0,1]. Then combine those two numbers
(either via addition or multiplication or max-function) and sort by that.
I believe this is the algorithm used by "real" news like, for example, CNN
Headline News. It lets something slightly older but with a higher score
outrank something slightly newer but with a lower score, but also lets
something newer with a high score outrank something older with the same score.
Thus, everything making it into that list starts out at some position and then
either goes up a bit (if more people come along and rate it higher) but always
is guaranteed to drop at some point. Super-duper-duper important news tends
to hover near the top much longer than "only" important news.
--Todd
|
|
| |
Todd Lehman wrote:
> Hmm...Maybe, but I meant fuzzy in the fuzzy-logic sense -- take the internal
> score (0-100) and convert it to a real number in the unit interval [0,1], then
> take the age of the message relative to the specified time period and make
> that also a number in the unit interval [0,1]. Then combine those two numbers
> (either via addition or multiplication or max-function) and sort by that.
> I believe this is the algorithm used by "real" news like, for example, CNN
> Headline News. It lets something slightly older but with a higher score
> outrank something slightly newer but with a lower score, but also lets
> something newer with a high score outrank something older with the same score.
> Thus, everything making it into that list starts out at some position and then
> either goes up a bit (if more people come along and rate it higher) but always
> is guaranteed to drop at some point. Super-duper-duper important news tends
> to hover near the top much longer than "only" important news.
nodnod, that's what I expect from the top40 page, but for the spotlight
(which I do like to think of as day oriented) the age of the message
shouldn't matter that much, cause the set we're ranking is (should be,
IMO) restricted to one day...
:)
Dan
|
|
|
