On 5/27/03, PowerMail discussions wrote:

>Subject: Re: Powermail 4.1.3  (Was: Powermail 4.1.2)
>From: "Rick Lecoat" <[EMAIL PROTECTED]>
>Date: Mon, 26 May 2003 20:36:18 +0100
 snip

>PM returned 3 messages. None of them contained the word "total" and none
>of them contained the word "carnage".
>Conclusion: PM's search function is still useless in the extreme because
>I will never be certain that it has found all of the matching messages,
>and in order to find those matching messages I will normally have to wade
>through many results that do not match any of my search terms. 

OK, but did ANY messages in your database have those words?  (you said
you picked some words at random).  If so, how many messages did it miss?

If PM missed messages that had one or both words, that is a serious
problem.  If you didnt have any messages with those words, then your
conclusion is invalid.

A "vector search" like PM uses will return a lot of hits that do not have
both words, but the ranking system should put the ones with both words
near the top of the list.  It should also do a tolerable job of finding
what you want even if you dont know EXACTLY what you are looking for.  In
my limited experience with PM (and a lot more experience with the Apple
V-twin search engine that it uses), it generally does a pretty good job
of that. (and way better and faster than the boolean search in eMailer).

Some of the problems are not in the basic search algorithm, but rather in
the "stemming", which attempts to reduce variants of a word (like
stemming, stems and stem) into a core word that then goes into the
search.  Given the nature of English, this is a tough problem (like mouse
and mice), and the messages with hits might have had a word that was
similar enough to your search words to show a hit.

One thing is obvious, people who are used to boolean searching tend to
not like the "fuzzy" results that come from vector searches.  Oh well.

bob

*************************************************************************
*  Bob Parks                           Time Flies Like The Wind,        *
*  [EMAIL PROTECTED]                   Fruit Flies Like Bananas.        *
*  [EMAIL PROTECTED]                                   Marx            *
*  http://www.kidsource.com/                                            *
*************************************************************************


Reply via email to