> 
> Looking in CVS Web I found this sql.c comment for v1.4 (ram) very suspicios:
> 
>         Document rating store in pos field of wrd[] = n_uniq_words
> 
> According  to  this, rating always will be 1 for one-word searches, since
> uniq word is always equal 1. Rating is totally meaningless in this case!
> Moreover,  for  many-words  searches  it  is  not  good  too,  sice can't
> distinguish  between (300 word1 + 300 word2) document and (word1 + word2)
> document.

 Document ordering takes it in account. Take a look into
doc/relevancy.txt
what factors take part in ordering. If you have an idea of how to
compose
function which can give document relevancy in persents, please share it.
It must take in account all factors explained in relevancy.txt. And
it also must give always the same persentage for the same document when
the same words are searched, does not matter whether tag, URL or other
limits  
are given. So, we can't just give 100% for the best document.


> Please reconsider this move.
>
___________________________________________
If you want to unsubscribe send "unsubscribe general"
to [EMAIL PROTECTED]

  • ... Andrey A. Chernov
    • ... Alexander Barkov
    • ... Andrey A. Chernov
    • ... Alexander Barkov
    • ... Андрей Чернов
      • ... Alexander Barkov

Reply via email to