clindh@Stride.COM (Christer Lindh) (08/22/88)
A *long* time ago I asked about free-text databases, and promised to
post a summary. Well, here it is:
-1-------------------------------------------------------------------------
I got a call from Ray Smith at Analythics. They have a VME-bus based
search-engine that is used by Sun abd Gould among others. I got a
demo-login to their machine, but haven't tried it yet. He said he
would send me some documentation but have not received anything yet.
-2-------------------------------------------------------------------------
From: johnl@ima.ISC.COM (John R. Levine)
The crufty old "refer" program included some programs that could run over
a large set of files, typically a whole directory tree, and index the first
100 "interesting" words (not in an ignore file) in each file. Then there
was a user utility where you typed in some words, and it told you what files
contained them in their hit lists. I gather these were, perhaps are, used
by the original clients of nroff, the patent application people at Bell Labs.
Technically, it used the same techniques as refer, hashed lists of the
interesting words, each truncated to 6 characters. Like the rest of refer,
it just barely worked and the code sent readers running for the dramamine.
-3-------------------------------------------------------------------------
From: mcgp1!fst
I know of one. AT&T ToolChest(R) has a system called UNITY, which works
on text based tables (which are easily workable with awk scripts as well).
You can have indexes on these tables. You also may want to check ZEBU(sp?)
or 'dbm'. I'm not sure if these two are text based. You can get the sources
for UNITY for around $1100.00 (This is last year's price).
UNITY is a collection of programs and a library of C routines. It requires
dictionary files (D files) which specify the fields for the table. These
field names are then used in queries, etc. It also has some nice routines
for data verification (i.e. matching a regular expression). A typical query
may look like this
$ uselect from /etc/passwd where loginid leq 'fst'
$ uselect from /etc/passwd where loginid req 'f.t'
NOTE: no field specification after 'uselect' means all of the fields in that
table. 'leq' means 'lexical equal'. 'req' means 'regular expression equal'
Skip Tavakkolian
..!uw-beaver!tikal!mcgp1!fst
-4-------------------------------------------------------------------------
From: schiltz@Sun.COM (Mara Schiltz)
Subject: re: Info on free-text databases wanted
Try BRS-search. I don't know how to get ahold of them, but
they certainly have what you're looking for and the run on Suns.
Mara Schiltz
-5-------------------------------------------------------------------------
From: Gary Puckering <cognos!garyp@RELAY.CS.NET>
Organization: Cognos Inc., Ottawa, Canada
We at Cognos started looking around for free-text database software
about 8 months ago and were suprised to find perhaps the best in the
world right in our own backyard. Try Fulcrum Technologies Inc.
They've licensed their software to Siemens, ICL, CPT (a
Minneapolis-based word-processing firm), LATICORP, COMTERM, VU/Text
(a divsion of Knight-Ridder), Library of Congress (used for on-line
catalog search), Hewlett-Packard (for their on-line CD-ROM manuals),
and many others. Fulcrum competes with BRS but is designed more as
a package for value-added resellers to integrate with their products
than as a standalone free-text system. Their address is:
Fulcrum Technologies Inc.
500 Rochester Street
Ottawa, Canada K1S 5K2
(613) 238-1761
Telex: 053-3046
-6-------------------------------------------------------------------------
From: roger@DPW
Organization: DP&W, New York, NY
Try getting in touch with Fulcrum Technologies -
560 Rochester Street
Ottawa Canada K1S 5K2
(613)238-1761
-7-------------------------------------------------------------------------
From: dartvax!daves@brspyr1.brs.com (Dave Schubmehl)
BRS/Search is a free text database management system that runs very well
under UNIX System V, 4.XBSD, Xenix, and several other variations as well.
Search takes documents as input, indexes every word, except for stopwords,
and allows you to search for combinations of those words, such as
"software adj engineering". Note: "adj" is our adjacency operator.
If you're interested, call BRS at 518-783-1161 and ask for the software
sales department. They can send you the marketing literature.
---------------------------------------------------------------------------
That's it. If someone has anything more to say on the subject, feel
free to mail me and I'll post another summary.
--
clindh@stride.COM :: Remember when we said there is
Stride Product Group, R&D :: no future ?
MicroSage Computer Systems, Inc. :: We're there now...
Reno, NV, (702) 322-6868 :: -Blank Reg