Re: bit of a history question

DaveT wrote:

> I was told by a 'consultant' that text searching in relational databases
> (like oracle) was not good in the early to mid 80's, so that's why the
> product they use works with a flat-file database instead of an RDBMS.
>
> Is this true?
>

Back in the early 80s it was very common to write flat-file (and card) based searches. These were based on the KWIC/KWOC index concept.

Effectively you has a "file" of Key Words Out of Context (KWOC) which was usually a master lis of words to ignore - the 'AND', 'OF', 'AS', 'THE' variety. One word per record.

And a "file" of Key Words In Context (KWIC) which represented words that were meaningful.

And a "file" of source document which was turned, by a program, into a one-word-per-record list.

And you ran the syncsort routine to eliminate the words out of context and count the words in context.

Since Syncsort was very efficient, this ran extremely fast and gave a list of documents and associated word counts. You then did stuff with the results manually. Not very efficient post-op processing.

That started to change with Oracle's SQL*TextRetrieval (and it's competitors) in the late 80s.

There is very little in common between KWIC/KWOC file-based search with what is possible today. The concept of a single SQL statement [that looks up email addresses (based on the author's name) for documents that contain references of specific words near others] was yet a pipe dream.

That concept has been implemented in Oracle's Intermedia and Oracle Text - at no additional charge for any edition of currently available databases.

As has [where the authors' address(es) is/are within a defines geographic radius from my location] through Intermedia Locator in that same SQL statement. Received on Tue Jul 13 2004 - 19:28:05 CDT