home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Simtel MSDOS 1992 December
/
simtel1292_SIMTEL_1292_Walnut_Creek.iso
/
msdos
/
hypertxt
/
hyperrez.arc
/
LINK26
< prev
next >
Wrap
Text File
|
1989-01-02
|
4KB
|
72 lines
Eight (8) problems with textual databases <sum10 1 9>
=========================================
Do search-driven textual databases -- such as Lexis, WestLaw,
Dialogue, etc. -- actually work? Here's why I ask the question:
Text-search programs violate all six (6) rules <link04> essential
to the building easily-accessed systems of knowledge. For example:
Problem (1) requires knowledge of the subject vocabulary in order to search,
======= (2) requires knowledge of the subject synonyms to broaden search,
(3) requires knowledge of Boolean operations for conducting searches,
(4) contains no capability for browsing (only verification),
(5) cannot insure completeness in finding all relevant material,
(7) typically retrieves irrelevant material with relevant material,
(8) shows citations instead of knowledge (ideas and relationships)
Fact To support the claim that textual databases contain fundamental
==== operational flaws, a recent study of the information retrieval showed
that lawyers using the nation' largest legal databases found on an
average only 22% of the relevant citations. What does this mean?
Hint You seldom win at five-card poker playing a one-card hand (22% info).
==== (press PgDn)
(press PgUp)
Shocking -- but even professional database searchers on DIALOG estimate
that they find less than 50% of the relevant material. Is that ok?
Retrieval of professional information
=====================================
Perhaps these low retrieval rates are useful for some academic work,
but in most professions, if your retrieval system misses even portions
of the relevant rules, specifications, procedures, diagnostics, or
drug interactions, you'll spend your life in court fighting a variety
of malpractice or liability suits. Question -- do text searches work?
Yes! (1) Text searches do work on familiar material (recall)
No! (2) Text searches don't work on unfamiliar material (comprehension)
This leads to the fundamental problem with on-line databases and
CD-ROMS. Regardless of the search software, with large bodies of
unfamiliar material, such retrieval systems having different results
for different users make retrieval an art rather than a science.
Furthermore, this art is not teachable, particularly as it assumes
knowledge of all synonym patterns used to express ideas.
(PgDn)
Fundamental problem (PgUp)
===================
The fundamental problem with textual databases (inverted word
indexes) is that successful retrieval in linguistic systems
ultimately expect from each user a vast knowledge of the world.
In contrast, success with hypertext retrieval systems is almost
independent of the abilities of each user (providing that the
builder of the system understands the world of knowledge <link60>).
While you may read about wonderful CD-ROMS databases (priced at from
$1,000 to $100,000 per year), don't be the first on your block to get
one. Why?
If the CD-ROM sellers expect you to organize the information (using
their database search routines), in many cases they are really
operating a garbage dump while charging outrageous fees to scavengers.
Maybe you can find your needle in their unorganized haystacks -- but
odds (supported by most experienced data retreival experts) say no.
Fact Boolean retrieval is an art typically removed from a dependable sceince
==== or process that most users need. This explains the growing interest in
hypertext whcih creates consistent and dependable access to knowledge.