Subject:
|
Top-N article list adjusted for article-age
|
Newsgroups:
|
lugnet.admin.general
|
Date:
|
Fri, 7 Apr 2000 06:47:51 GMT
|
Highlighted:
|
!
(details)
|
Viewed:
|
966 times
|
| |
|
|
In lugnet.admin.general, Todd Lehman writes:
> [...] The elapsed time (article age) will be accounted for later in
> searches and top-N lists by scaling the score according to some f(t). [...]
Here's a crude example -- a static list built "by hand" from a few one-liners
using current data:
http://www.lugnet.com/?p=hotlight
I told it to examine all the articles which have been rated so far, and
adjust their ratings downward the older the article, according to this
formula:
r' = r / (1 + t)
where r is the article's current rating, r' is the adjusted rating, and t is
the elapsed time (article's age) in weeks, yielding a smooth hyperbolic curve.
Thus, a brand-new article which has just appeared has its rating divided by 1
(no change). An article 1 hour old is has its rating divided by 1.006. An
article 1 day old has its rating divided by 1.1429. An article 3 days old has
its rating divided by 1.4286. An article 7 days old has its rating divided
by 2. etc., etc.
A more concrete example: the first article currently shown at the URL above
(if you're reading this way in the future from now, it might not match what
you read here!) currently has a rating of 83.3333 and was approximately 10
hours and 45 minutes (or approximately 0.06408 weeks) old at the time the
list was generated. Thus it had its rating divided by approximately 1.06408,
giving an age-adjusted rating of 78.2834.
The list shows all articles in the system with an age-adjusted rating of 60
or higher.
But it's just a test, and it's past my bedtime. The purpose of this test
was to see how closely the numbers might come to a "what's hot" list by
applying a simple age formula.
--Todd
|
|
Message has 2 Replies:
 | | Re: Top-N article list adjusted for article-age
|
| (...) that's pretty cool - what would be interesting is to see what articles were _missed_. Also, I started thinking about the default rating... Say, something posted to .announce should probably be rated higher by default than, say, something (...) (24 years ago, 7-Apr-00, to lugnet.admin.general)
|
Message is in Reply To:
| | Re: New 9V Digital Trains for Germany this Autumn
|
| (...) There are many different methods that people can use in their minds as they give input, but all of the methods have the same net effect: They give a number which gets averaged, and that average score is a reflection of a collective "read (...) (25 years ago, 5-Apr-00, to lugnet.admin.general)
|
24 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
