From: John Conover <john@email.johncon.com>
Subject: Enterprise wide information retrieval system
Date: Fri, 19 May 95 01:28 PDT
Attached is a description of a enterprise wide information retrieval
and documentation system (eg., "memo machine," "information robot,"
"infobot," or "mailbot,") that I have used with good results for many
years. It has been used for context management of projects, in
geographically, and ethnically disperse organizations, that are
involved in the very dynamic electronic component market,
(microprocessor market, to be more specific.) The "C" sources to the
system are available at no charge by sending an email with the
following format:
To: info-request@email.johncon.com
Subject: archive
get rel
The sources constitute about 400K bytes, and will be returned to you,
by return email, in two sections. (You will be mailing into the system
described below, BTW, only in this case, it is functioning as a source
repository-so the search functions are disabled.)
You may want to skip the 5 numbered paragraphs, below, to get to the
description of the conceptual application of the system. There are
several attachments that describe the technical evolution of the
system. I think it might be interesting reading for those involved in
modern organizational theory and informatics-particularly if the
market place served is very dynamic, (the revenue's of the
microprocessor industry have a time series that is fractal, BTW.) It
should be apparent that this system is not meant to replace the
traditional content database systems used in MIS organizations, but is
complementary to them. The system is capable of operating in a
distributed, interoperable client-server environment.
The second paragraph of numbered paragraph 5, below, concerning the
concept of "moving orthogonal in information space" may be of interest
to those involved in OD when doing organizational effectiveness
evaluations.
john@email.johncon.com (John Conover)
--
John Conover, john@email.johncon.com, http://www.johncon.com/
______________________________________________________________________________
The objective of this application of the rel program, in conjunction
with the procmail/smartlist programs, is to construct an enterprise
wide, full text information retrieval system that uses the Unix MTA
(Message Transfer Agent,) as a delivery, query, and distribution
system-in a sense, what is currently termed "groupware."
In general, the operational procedure is as follows:
1) The information retrieval database is installed in the
database administrator's, (perhaps a project team leader, since
installation does not require sysadm privileges,) account, as
per the the instructions in the INSTALL file in this
directory. Multiple information retrieval systems may be
installed, if necessary.
2) All members, perhaps a project team, who will have access to
the information system will be entered into the accept and/or
reject files, as appropriate, in the smartlist directory for
the system's database as per the standard installation in the
procmail/smartlist's manual. If the repository is, in addition,
to function as a distribution agent of information, the dist
file in the smartlist directory should include the email
addresses of those on the distribution list, as per the
standard installation outlined in the manual.
3) As formal email correspondence occurs between the members of
the project team, any mail that is mailed to the information
system's account will be saved, and placed under maintenence of
the full text database system. Anything pertaining to the
project should be submitted to the database system. If the
system functions as a distribution agent, all email submitted
will also be mailed to the distribution list, which could be
other team members, management, etc. These in turn, can be
replied to, and the replies will be placed under the
maintenance of the database system, and distributed to all
others on the distribution list-ie., it is an asynchronous
conferencing system, or an electronic "mailing list."
4) However, it differs from conventional electronic mailing
lists, in that it has a "sister account," which can receive
mail, and can be used to query the database for previous email
concerning issues that have been addressed, (or, perhaps not
addressed.) The program rel is used for the query, so a
"context can be framed," using the rel search criteria. Any
email found that concerns the "framed context," is returned to
the user via email, as a MIME compliant digest. The order of
the emails in the digest are in order of relevance to the
"context" of the query. In some sense, the digest is like an
electronic administrative folder that is maintained with
correspondences concerning specific issues. The major
difference is that the folder can be created dynamically, on
demand.
5) It turns out, that with a modern MUI (Mail User Interface,)
like the GNU emacs' rmail facility, that very powerful
informatic operations can be performed. For example, the email
digest returned from the operation in 4) above is in a MIME
digest, in order of relevance to a "context" query. The digest
can be "burst" into individual messages, and, starting with the
most relevant message, the messages can be browsed to find a
specific entry, or the context of group of messages. If a
specific message is of interest, it can be replied to, to
re-table an issue, or request additional information,
etc.
There is one other very powerful operation that can be
performed. Suppose that when examining the messages in order of
relevance, some interest is generated in one of the messages,
and it is desired to find out more about the "context" of the
message. The messages can be rearranged, and ordered by the
author to further find out the the "context" of the messages
from the author's point of view. The messages can also be
rearranged by date, so that a short "context" window can be
derived around the temporal issues of the message. These
operations are termed "moving orthogonally in information
space," eg., you were moving through the documents in order of
relevance, then investigated the concerns of a specific author,
then investigated the author's concerns in relation to the rest
of the group over a period of time, etc.
Note that the process outlined in sections 3), 4), and 5), above,
really constitutes nothing more than an electronic literature search,
and is a similar process to what a historian would do when researching
a subject, (presumably because an understanding of the "context" of
the subject was desired.) The only difference is that the process is
highly automated. The proposed system can be thought of as an
electronic filling cabinet that can be searched, electronically. The
concept is not new-it was proposed by Vanavar Bush in the 1940's, (the
Memex Machine,) and later modified by Douglas Engelbart in the early
1970's.
Note that what is being proposed is a new administrative paradigm-one
that addresses context as opposed to traditional content issues in
organizations, and is compatible with the contemporary concepts of
"Empowerment," and "Total Quality Management." Administration is the
mechanization of the flow information through an organization. What is
being proposed is to use an "information machine," ie., computer, to
search, collate, and distribute information. In some sense, it is a
"memo machine" that can transcend organizational and parochial
boundaries. Note that the memos do not have to be structured-how
information is structured will be specified at query time, not the
time of composition of the memo. (Note that it is not a "hypertext"
system, since the "links" are constructed, dynamically, at the time of
the query-not at the time a document is composed.)
A key issue in automating the process is the capability for the
uninitiated to create "context" queries that are representative of the
information desired-the rel syntax is powerful, intuitive, and easy to
use on an operational basis; most managers already understand how to
use email, and the rel query syntax is similar to the one used in
algebraic calculators-in point of fact, it is identical to the syntax
used in calculators from Texas Instruments and Casio, except that
numbers are replaced by words, and mathematical operators are replaced
by boolean operators-a short hand natural language query.
As a concluding remark, note that the system can not be a substitute
for good organizational practices and disciplines-as Doug Engelbart
stated "if you automate a big mess, you end up with a very fast big
mess."
There are 3 attachments. The first is a brief introduction of how the
system was used in an experimental program, the second is excerpts
from the original system development reports, (the system outlined,
above, is far simpler than the original,) and the third is an excerpt
from the rel program manual.
______________________________________________________________________________
Attachment 1:
______________________________________________________________________________
Attached is a brief synopsis of an asynchronous conferencing system
(also known as an information retrieval system, electronic literature
search system, or corporate repository,) that I used in cross
functional program management, in another life, a long time ago. The
objective was to find a methodology to relate the corporate
information repository to the management structure, (we did not
consider the technical issues to be significant.) The general concept
was to add sufficient functionality to the Unix email system to turn
it into an electronic literature search system.
The attached is a "cut and stick" from some of the reports on the
system's development. The project/program team supported by this
system consisted of little over a hundred professionals, from
approximately 20 specialties, and 4 core corporate functions. They
were geographically, and ethnically, disperse.
Also, attached is the description a program, rel.c, which was used to
perform complex literature search queries on the full text database,
and return documents, in order of relevance to the query, (note that
hypertext methodologies are incapable of operating in this fashion.)
These documents were returned as an email digest, which could be
"burst," into constituent documents, allowing the most relevant
document to be reviewed first, (and then, if necessary, the specific
email could be responded to, for further clarification, etc.) Since
most email readers, elm, pine, email, etc., are capable of sorting a
mail box by different criteria, one could move "orthogonally in
information space," during the review process, (ie., move though the
documents by relevance, and then sort by author, to find out what
he/she had to say about things, then sort by date, to find out the
chronology of events in the discussion, etc.) The database system was
a distributed environment, with each segment of the database
consisting of less than 10Mbyte, so queries were done in parallel,
using all network resources, and, thus, very fast.
In point of fact, all of the attachments are a "cut and stick" from
documents fetched from the full text database system with command "rel
((information & retrieval) | (literature & search) | (corporate &
repository) & management)"
______________________________________________________________________________
Attachment 2:
______________________________________________________________________________
Various "cut and sticks" from the development reports.
Information systems are used in program management, which must
coordinate the various activities of the corporate functions (ie.,
engineering, marketing, sales, etc.) involved in development
projects. After researching the issues, (see below,) We concluded that
a distributed full text system that uses the mail (MTA) system as a
communication medium is the desirable direction to pursue. Our
reasoning is as follows:
1) The Unix MTA is almost universal, and will operate
effectively over uucp and/or ethernet connectivities in a
non-homogeneous hardware environment.
2) Each transaction is logged, with a date/time stamp, and who
created the transaction.
3) The MTA already has remedial file storage capabilities,
which can be used to query/respond to transactions at a later
date.
4) Most(?) computers are already connected together, and users
are familiar with how to use the system.
5) The MTA database can be NFS'ed to conserve machine
resources.
6) It is a text based system.
We discounted the "hyper text" type of systems, because the links must be
established before the document is stored-which is fine if you know what
you are going to query for. In a general management application, this is
seldom the case. We set up a prototype system, using the following
(readily available) programs:
1) elm, because it has a slightly more sophisticated file
storage structure, and a very powerful aliasing capability
that can alias team members as a group. Additionally, it has
limited query capabilities, and can, through its forms
capabilities, send mail transactions in a structured format.
(Which is advantageous if the transactions are used for
notification of schedule milestone completion, etc.) Eudora
was used on the PC's and MAC's, using POP3 as the
communications environment between the PC's and the Unix MTA.
2) The dbm library to build an extensible hash query system
into the file storage structure made by elm. This was
operated in two ways, by an RPC direct call, and a mail daemon
that "read" incoming mail (to a query "account") and returned
(via mail) all transactions that satisfied boolean
conditionals on requested words. (A data dictionary was added
later, so that the dictionary could be scanned for matches to
regular expressions, which were then passed to the extensible
hash system, but for some reason, this was seldom used.) The
query was made through a very simple natural language
interface, ie.,
send john and c.*r not January
would return all transactions containing john, excepting those
written in January. (We did not attempt phrases, it looked
complicated-this is ill advised by Tenopir, etc. below.)
This program contained approximately 350 lines of C code. A
soundex algorithm was added later to overcome spelling
errors-the full text database contained the soundex of the
words in a document, and any words searched for were converted
to soundex prior to the query. (See the works by Knuth for
details of the soundex algorithm.) Also a parser was added so
that the boolean search words could be grouped in postfix
expressions, eg., ((john & conover) ! (January | march)). The
order that the documents were returned in is in order of
relevance.
This prototype was well received, and was used as follows:
1) Management "decreed" that the system would be used as a
management tool, and all data had to be entered, or
transcribed into the system (including the minutes of
meetings, etc.) If it didn't exist in the system, it did not
exist. All discussions, and reasons for decisions had to be
placed in the system. ALL team members and upper management
had identical access to ALL transactions. (Mail could be used
for private correspondence, such as politicking, etc. but all
decisions, and the reasons for the decisions had to be placed
in the system.) The guiding rule was that at the end of the
project, the system contained a complete play by play
chronology and history of all decisions, and reasoning
concerning the project, and, by the way, who was responsible
for the decisions. On each Monday, everyone entered into the
system, his/her objectives for the week, and when each
objective was finished, she/he mailed the milestone into the
system-ie., all group members and management could thus find
out the exact status of the project at any time (ie., a
"social contract" was made with management and the rest of the
members of the team.) In some sense, it is really nothing more
than an automated, real-time MBO system. At any time, a
discussion could be initiated on problems/decisions in the
system by anyone. The project manager was assigned the
responsibility of "moderator," or chair person for his/her
section of the project. Each Friday, the system was queried
for project status, and the status plumbed to TeX for
formating, and printed for official documentation. This
document was discussed at a late Friday people-to-people staff
meeting. (The reason for setting things up this way can be
found in Davido, below.)
2) Marketing was responsible for acquiring all market data on
magnetic media, (from services like Data Quest, the Department
of Commerce, etc.) and each document was "mailed" into the
system so that the information was available for retrieval by
anyone. All had access to the progress made by engineering,
and can contribute information on issues as the program
develops-ie., this was a "concurrent engineering" environment.
3) Engineering was responsible for maintaining schedules, and
reflecting those schedules in the system-if slippages occurred
the situation could be addressed immediately by management,
and a suitable cross functional resolution could be arrived
at.
4) Sales was responsible for adding customer inputs,
concerning the project, into the system, so customer
definitions could be retrieved by all project members. This
included the customer data, such as who has buying authority
in the customer's organization, who has signature, etc.
The results were very impressive not only by productivity standards, but
also by "correctness to fit and form" standards (ie., the right product
was in the market at the right time, the first time.) This has becoming a
central agenda, as outlined in Davido, below.
Bibliography:
"Computer-Supported Cooperative Work," Irene Greif
"A model for Distributed Campus Computing," George A. Champine
"Enterprise Networking," Ray Grenier and George Metes
"Connections," Lee Sproull and Sara Kiesler
"5th Generation Management," Charlse M. Savage
"Intellectual Teamwork," Jolene Galegher, Robert E. Krout and Carmen Egido
"In the Age of the Smart Machine," Shoshana Zuboff
"The Virtual Corporation," William H. Davido and Michael S. Malone
"Accelerating Innovation," Marvin L. Patterson
"Paradigm Shift," Don Tapscott and Art Caston
"Developing Products in Half the Time," Preston G. Smith and Donald G. Reinertsen
"Full Text Databases," Carol Tenopir and Jung Soon Ro
"Text and Context," Susan Jones
______________________________________________________________________________
Attachment 3:
______________________________________________________________________________
Rel is a program that determines the relevance of text documents to a
set of keywords expressed in boolean infix notation. The list of file
names that are relevant are printed to the standard output, in order
of relevance.
For example, the command:
rel "(directory & listing)" /usr/share/man/cat1
(ie., find the relevance of all files that contain both of the words
"directory" and "listing" in the catman directory) will list 21 files,
out of the 782 catman files, of which "ls.1" is the fifth most
relevant-meaning that to find the command that lists directories in a
Unix system, the "literature search" was cut from 359 to 5 files, or a
reduction of approximately 98%. The command took 1 minute and 26
seconds to execute on a on a System V, rel. 4.2 machine, (20Mhz 386
with an 18ms. ESDI drive,) which is a considerable expediency in
relation to browsing through the files in the directory since ls.1 is
the 359'th file in the directory. Although this example is remedial, a
similar expediency can be demonstrated in searching for documents in
email repositories and text archives.
General description of the program:
This program is an experiment to evaluate using infix boolean
operations as a heuristic to determine the relevance of text files in
electronic literature searches. The operators supported are, "&" for
logical "and," "|" for logical "or," and "!" for logical "not."
Parenthesis are used as grouping operators, and "partial key" searches
are fully supported, (meaning that the words can be abbreviated.) For
example, the command:
rel "(((these & those) | (them & us)) ! we)" file1 file2 ...
would print a list of filenames that contain either the words "these"
and "those", or "them" and "us", but doesn't contain the word "we"
>from the list of filenames, file1, file2, ... The order of the printed
file names is in order of relevance, where relevance is determined by
the number of incidences of the words "these", "those", "them", and
"us", in each file. The general concept is to "narrow down" the number
of files to be browsed when doing electronic literature searches for
specific words and phrases in a group of files using a command similar
to:
more `rel "(((these & those) | (them & us)) ! we)" file1 file2`
Although regular expressions were supported in the prototype versions
of the program, the capability was removed in the release versions for
reasons of syntactical formality, for example, the command:
rel "((john & conover) & (joh.*over))" files
has a logical contradiction since the first group specifies all files
which contain "john" anyplace and "conover" anyplace in files, and the
second grouping specifies all files that contain "john" followed by
"conover". If the last group of operators takes precedence, the first
is redundant. Additionally, it is not clear whether wild card
expressions should span the scope multiple records in a literature
search, (which the first group of operators in this example does,) or
exactly what a wild card expression that spans multiple records means,
ie., how many records are to be spanned, without writing a string of
EOL's in the infix expression. Since the two groups of operators in
this example are very close, operationally, (at least for practical
purposes,) it was decided that support of regular expressions should
be abandoned, and such operations left to the grep(1) suite.
Applicability:
Applicability of rel varies on complexity of search, size of database,
speed of host environment, etc., however, as some general guidelines:
1) For text files with a total size of less than 5 MB, rel, and
standard egrep(1) queries of the text files will probably prove
adequate.
2) For text files with a total size of 5 MB to 50 MB, qt seems
adequate for most queries. The significant issue is that, although
the retrieval execution times are probably adequate with qt, the
database write times are not impressive. Qt is listed in "Related
information retrieval software:," below.
3) For text files with a total size that is larger than 50 MB, or
where concurrency is an issue, it would be appropriate to consider
one of the other alternatives listed in "Related information
retrieval software:," below.
References:
1) "Information Retrieval, Data Structures & Algorithms," William
B. Frakes, Ricardo Baeza-Yates, Editors, Prentice Hall, Englewood
Cliffs, New Jersey 07632, 1992, ISBN 0-13-463837-9.
The sources for the many of the algorithms presented in 1) are
available by ftp, ftp.vt.edu:/pub/reuse/ircode.tar.Z
2) "Text Information Retrieval Systems," Charles T. Meadow,
Academic Press, Inc, San Diego, 1992, ISBN 0-12-487410-X.
3) "Full Text Databases," Carol Tenopir, Jung Soon Ro, Greenwood
Press, New York, 1990, ISBN 0-313-26303-5.
4) "Text and Context, Document Processing and Storage," Susan
Jones, Springer-Verlag, New York, 1991, ISBN 0-387-19604-8.
5) ftp think.com:/wais/wais-corporate-paper.text
6) ftp cs.toronto.edu:/pub/lq-text.README.1.10
Related information retrieval software:
1) Wais, available by ftp, think.com:/wais/wais-8-b5.1.tar.Z.
2) Lq-text, available by ftp,
cs.toronto.edu:/pub/lq-text1.10.tar.Z.
3) Qt, available by ftp,
ftp.uu.net:/usenet/comp.sources/unix/volume27.
______________________________________________________________________________