Re: Suggestions for adaptive-word-scoring

Re: Suggestions for adaptive-word-scoring

[Robert Bihlmeyer]

Hi,

[forwarding to the list on Lars's request]

>>>>> On 03 Sep 1996 04:59:22 +0200
>>>>> gnus-bug@ifi.uio.no (The Gnus Bugfixing Girls + Boys) said:

 Lars> Robert Bihlmeyer <robbe@orcus.ping.at> writes:
 >> I think it would be more logical to divide the score for an
 >> article by the number of words (in the subject) scored before
 >> scoring them, since more words in the subject mean a smaller
 >> "contribution" of each one to the worthiness of the whole.
 >> Otherwise (if scoring each word by a fixed amount) one would have
 >> articles with longer subjects going faster to hell or heaven.

 Lars> Yes, but wouldn't you want an article with the subject "Gnus
 Lars> Gnus Gnus" to get a higher score than an article with "Gnus"? I
 Lars> don't know. Could you make this suggestion to the
 Lars> ding@ifi.uio.no mailing list and see what people think?

If we really want to get a higher score for "Gnus Gnus Gnus", a
solution would be to modify the score each word in the subject (modulo
ignored ones) once for each appearance. So with a base score of +30 for
"read", a read article with subject "Gnus, September Gnus" would yield
+10 for each word, and finally +20 for "Gnus" (two appearances) and
+10 for "September".

Please CC to me, because I'm not on the list,

	Robbe