Subject | Re: [Firebird-Architect] Re: Google-like scoring in databases |
---|---|
Author | Jim Starkey |
Post date | 2003-06-30T15:38:29Z |
Roman Rokytskyy wrote:
more fields of the
same record and hits in records related by a primary/foreign key
relationship. Netfrastructure
recognized the first and has architectural provision for the second.
In practice, however, a database, even a gigantic database, is trivial
next to almightly Google.
And, given the nature of database applications, the scope of searching
is likely to be much
for directed. On a municipal site, for example, somebody is much for
likely to search for
"dog licenses" than "what is the meaning of life". I think things like
root searches, plural
handling, spell correcting are much more important than scoring links.
Or, to put it
another way, I haven't had any complaints in that direction.
But let me throw it open. Most of the people on the list have probably
been into the
IBPhoenix knowledge base which is implemented in Netfrastructure. How
would you
rate the hit scoring on that site? Have you been able to find what you
were looking for
reasonable near the top of the search list?
>Hower, in addition to words, phrases, etc. St.Google uses someBut a database has some things not available to Google: hits in two or
>meta-information that is not available in the document itself:
>directed relations between documents. This meta-information is added
>manually by human to link one document to another (except Wiki). And
>exactly this meta-information is believed to make Google results far
>more relevant than results of other search engines.
>
>And exactly absense of this information in the database does not let
>me relax. We loose this information when we map object (document)
>model of the application to the relations. And this leads me to belief
>that without introducing some mechanism to keep this information in
>database, we will not be able to implement search in Firebird at all
>(or, more correct, implementation will produce too much garbage).
>
more fields of the
same record and hits in records related by a primary/foreign key
relationship. Netfrastructure
recognized the first and has architectural provision for the second.
In practice, however, a database, even a gigantic database, is trivial
next to almightly Google.
And, given the nature of database applications, the scope of searching
is likely to be much
for directed. On a municipal site, for example, somebody is much for
likely to search for
"dog licenses" than "what is the meaning of life". I think things like
root searches, plural
handling, spell correcting are much more important than scoring links.
Or, to put it
another way, I haven't had any complaints in that direction.
But let me throw it open. Most of the people on the list have probably
been into the
IBPhoenix knowledge base which is implemented in Netfrastructure. How
would you
rate the hit scoring on that site? Have you been able to find what you
were looking for
reasonable near the top of the search list?