At 11:40 AM -0800 11/25/99, Aaron Turner wrote:
>Hmmmm. I guess I'll have to look into what you're doing. I didn't give
>much thought into doing it that way because I figured it would be too CPU
>intensive to search that way. My solution after about 30 minutes of
Nah. You have an inverted index of words, each pointing to a record
with the appropriate location information. So a phrase search
essentially starts out as a boolean AND, then filter from there.
For your purposes, you'd be doing whatever boolean type the user
wanted, but then you'd weight the words in the query based on
location. Simply run through the generated word list once and add the
appropriate scoring.
>have 10 columns, each column holds the words 1 to 10 away. The following
>sentance would have the following entries in the table:
That certainly wouldn't have been my first thought. I can see why
you're using up so much disk space. I just don't see how this is
going to be faster than an inverted index... You're still adding an
entry in the DB for every word (so there aren't any fewer entries
than the current approach), plus each entry seems like it'd be a lot
larger.
-Geoff
------------------------------------
To unsubscribe from the htdig3-dev mailing list, send a message to
[EMAIL PROTECTED]
You'll receive a message confirming the unsubscription.
