Professional Documents
Culture Documents
Selective Dissemination of Information (Sdi) : State of The Art in May, 1963
Selective Dissemination of Information (Sdi) : State of The Art in May, 1963
Selective Dissemination of Information (Sdi) : State of The Art in May, 1963
257
258 PROCEEDINGS—SPRING JOINT COMPUTER CONFERENCE, 1963
is well under way,11 The Douglas Aircraft Cor- months effort would be necessary to install an
poration in Santa Monica, California, has a early SDI system.3 Programmers have been
system in an advanced state of debug.12-13 used in all cases. The time to get a SDI pro-
Around the end of 1962, a second 1401 program gram through a monitor system or to fit in with
became operational at IBM, Data Processing other existing operating procedures has been
Division, Midwest Region, in Chicago, Illinois. quite variable. In one case the program as-
Over the past four years, SDI has moved from sumed a particular load routine long in general
a concept into a rapidly increasing number of use, but not in use in this installation.
system installations. Experience with combining and modifying
existing systems is exemplified by Poughkeep-
Implementation sie. There, despite the fact that the programs
Implementation difficulties for a system are had little, if any, documentation, one or two
often underestimated. This is in contrast to programmers fought through SDI, KWIC* and
reduced operational cost and increased quality an IR program in a few months. The manual
of service which are often overestimated. With procedures were in flux for a longer period.
SDI, the choice is whether (1) to use an avail- The total system is still being modified and only
able system, (2) modify an available system, or parts are in operation. Owego was a rewrite
(3) to write your own new one. Because of the from SDI 2 which took over a calendar year to
uncertainty, implementation cost is hard to get into operation. The program rewrite itself,
estimate. Quality is even more difficult to esti- from start to run, took about three months.
mate. Everyone seems to feel he is an expert Prestart systems work extended longer and, to
on quality. There is disagreement in many my knowledge, the system is still rather weakly
cases. Present SDI systems involve computer documented for remote installation 18 ' 19 and is
programs, manual procedures and sometimes being integrated with KWIC.7 What might
special equipment. In order of increasing diffi- seem to be a relatively simple rewrite of SDI 2,
culty, implementation may involve the installa- SDI 3, required one person for a calendar year
tion of a well-documented, tried and true sys- in a building being noisily rebuilt, although the
tem which is in operation somewhere else; programming and documentation was done by
modification of manual procedures; obtaining an experienced programmer who knew SDI and
special equipment; reprogramming or redesign. the machine.
Human skills available; experience of the The classic problem seems to be an under-
personnel with SDI, Information Retrieval or estimation of the amount of the programming
related areas; and the number of other systems, required to rewrite and document. For experi-
procedures or constraints interacting with the enced personnel, e.g. SDI 3, estimates seem to
new SDI installation all affect the effort re- be low by a factor of four. For less experienced
quired for implementation. Not only are a wide (with SDI) personnel perhaps six would be
systems background, computer knowledge, and better, e.g., Poughkeepsie. It should be pointed
documentation experience valuable, but spe- out that certain phases can sometimes be esti-
cialized knowledge with office machinery, in- mated accurately, e.g., programming at Owego.
dustrial engineering, typography as well as
psychology, sociology and organization theory User Interests
often help. Programmers seem to be necessary Most user profiles (interests) have been ob-
for any type of installation. The more experi- tained without any problem by blindly mailing
ence with data processing as contrasted with a short form to the potential user.t In three
scientific programming the better, but any pro- tests* some 65% of those contacted became
gramming experience is better than none. Sys- users. Mass meetings of potential users have
tems and procedures personnel are well-known been used as well as blindly mailing longer
in most organizations and are certainly advan-
tageous for modifications or rewriting.
t Key Words in Context, a machine prepared printed
Experience with installations of documented index.17
SDI systems is limited. It was estimated that * 5, Pages 94-5,
three calendar months and a total of three man t *, Page 41.
SELECTIVE DISSEMINATION OP INFORMATION (SDl) : STATE OF THE ART IN MAY, 1963 259
forms with either term dictionaries attached, on which there is experience is quite wide, in-
e.g., Owego (modified ASTIA), or enclosing cluding science, engineering and management.
examples*** of indexed document items. In- There are no known cases of letters, memo-
direct methods have also been used to derive randums, or picture annotations being processed
profiles from personnel or project informa- although this has been proposed and no prob-
tion.*** Only with SDI 1 was a comparative lems are anticipated. Document source has been
study made and it had too small a sample to shown to be a significant factor in response.14
be conclusive.*** Each of these methods have Owego uses ASTIA documents predominately.
been proven feasible. Further research is Poughkeepsie uses internal IBM reports. Sur-
needed to define situations where one is pref- veys of what users read15 or library usage could
erable to another. be used to determine what document sources to
Adjustment of user profiles has been done use for an SDI system. Most such data indi-
largely at the user's instigation. At Mohansic, cates a skew distribution of usage with a few
blanket mailings of current user profiles with highly used journals. It is assumed but not
change forms have been made to encourage demonstrated that different types of users need
users to make changes. Users have also been different document sources. Experimentation
notified that they can make changes. The effec- in this area might influence the selection by
tiveness of these measures is subject to doubt. professional journals of items to abstract. SDI
The only known attempt to automatically up- provides a tool in this area through its response.
date or adjust profiles based on user's responses SDI 4 and a revised Chicago system will allow
was tried at Mohansic on SDI 1. The results exclusion of documents by source, e.g. need-to-
were inconclusive. Manual attempts to suggest know or excluding journals user subscribes to.
or arbitrarily make changes in user profiles Volumes of document items being processed in
based on various hypotheses have been made SDI systems run from tens to hundreds per day
from time to time, usually without controls. with experience upward lacking. Subscribing
Although how to get new users to join and to a journal is not much of a problem, but
give the "best" possible profile seems to be a getting on internal distribution lists is more
difficult theoretical problem, in practice there difficult than one might expect. It cost the
seems to be no difficulty. Experiments with Mohansic group several man months of effort
automatic updating are in order but adequate to locate internal sources of information and
user response histories seem to be necessary. arrange to be added to these distribution lists.
The number of users serviced by SDI systems Documents normally come to one location,
now in operation has ranged from tens, to are handled and numbered. Some SDI's inte-
one to two thousands. Experience with larger grate with library operations to various de-
groups is lacking although no new problems grees. Owego uses the same numbering and
are anticipated. One problem, not initially an- hard copy reproduction procedures. Mohansic
ticipated, which increasing number of users provides abstract sets and utilizes journals
has proven to be important, is that of address from the library. Douglas is partially inte-
changes. These occur so frequently that not grated. Some work with IR systems, e.g., Evan-
only must they be considered part of every dale, Owego, Washington. Document number-
normal run, but provision is necessary to ing may be sequential as at Mohansic or by an
rVlJmCTA p r i r l r ' o a o a c h a i i i T a c m n n f i f i n o + i r v - n rnnA \\n-vA int°niQ^ nnAa 0,3 Q+ QwgnrQ and Poughkeepsie.
copy order. As we shall see below (Abstracts Checking for duplication and series complete-
and Notifications) this affects the notification ness is a normal library problem.
itself. There seem to be few serious operational
problems in this area. Studies are needed to
Documents
test automatic procedures to analyze user re-
Document sources for SDI are usually defined sponses and to vary the document source mix
by the application. The range of subject matter to maximize value functions. Little has been
done to study the effect of frequency of mail-
*** Ibid., Page 6. ings to the user.
260 PROCEEDINGS—SPRING JOINT COMPUTER CONFERENCE, 1963
should be in appropriate sequential order for There have been various hard copy proce-
mailing. If 5, 6 and 7 are not machinable on dures. (1) Ignore the problem (SDI 2-4). (2)
return, response handling for document hard Refer the user to a library (SDI 2-4). (3)
copy orders and operating statistics must be Shelve and pull (SDI 1-4). (4) Keep vellum
manual, as in SDI 1. The abstract, 1, should and reproduce (SDI 2-4). (5) Use aperture
be retainable by the user. A study15 in one cards and reproduce (initially at Owego).
organization shows 3 x 5 and IBM card sizes (6) Use reel microfilm at multiple locations
were the most frequently used media for this (Poughkeepsie). Adequate analysis of cost
purpose even prior to SDI. The response, 5, is and value are yet to be made. Most systems
made at many remote uncontrollable locations. agree with the Mohansic survey,15 users want
The PORT-A-PUNCH® card has proven to to be able to obtain hard copy.
provide a machine readable response. PORT- Value-Cost
A-PUNCH is only now (February 15, 1963)
becoming available in continuous forms, thus This is, in my opinion, the area with the
making machine (1403) printing of the ab- largest potential for development. Available
stracts on the PORT-A-PUNCH card or an cost data 3 is very limited and hard to inter-
attached form possible. Previously a bill feed pret. Available value information15 is largely
attachment was necessary which slows the subjective. Dichotomy scales have been used
printer. Systems remain to be developed and in SDI, i.e., "of interest vs. not of interest."
tested based on bill feeds, optical reading and It is my opinion that ordinal, and cardinal
many other devices. When several notices go scales are needed if we hope to move SDI
to each user at once, placing several cards design from an art towards a science.
together (or using a sheet of paper as at
Evandale) might save handling expense and REFERENCES
user exasperation. No existing system meets 1. "A Business Intelligence System," H. P.
all of these requirements; each compromises LUHN, IBM Journal of R&D, 2, 4, 314-319,
to some extent. Considerable research is neces- October 1958.
sary before sufficient basic knowledge is ob- 2. "Selective Dissemination of New Scientific
tained as to the relative worth of these various Information with the Aid of Electronic
features. Processing Equipment," H. P. LUHN,
American Documentation, 12, 2, 131-138,
Response, Reports and Hard Copy April 1961.
SDI 2-4 require the user to respond on every 3. "Selective Dissemination—Report on a
notice. Other systems require responses under Pilot Study—SDI 1 System," C. B. HENS-
certain conditions (SDI 1, no response if nega- LEY, T. R. SAVAGE, A. J. SOWARBY, and A.
tive) or never, i.e., just a notification. It is RESNICK, IBM, ASDD, Yorktown Heights,
not known exactly what effects this has. N. Y. Report 17-039, January 1961. (Pre-
Responses and other records allow reports to sented at the 18th meeting of the Opera-
the user, operators, management and research tions Research Society of America—1960),
personnel. This is a largely undeveloped area 45 pp.
even though some rudimentary reports are in- 4. "Selective Dissemination of Information—
cluded in the SDI 2 and 3 systems* Feedback A New Approach to Effective Communica-
reports could be used to assist in updating user tion," C. B. HENSLEY, T. R. SAVAGE, A. J.
profiles, changing the document sources mix, SOWARBY, and A. RESNICK, IRE Transac-
adjusting the system sizes, changing indexing tions on Engineering Management, EM-9,
methods, and adjusting the cost vs. value bal- 2, 55-65, June 1962, 11 pp.
ance. Randomly selected notices (SDI 1-4) 5. "Selective Dissemination of Information—
allow the system selection performance to be SDI 2 System," W. BRANDENBERG, H. C.
compared to random selection as a base. This FALLON, C. B. HENSLEY, T. R. SAVAGE,
also allows miss items (which could have been and A. J. SOWARBY, IBM, ASDD, York-
selected by the system but were not) to be town Heights, N. Y. Report 17-031, April
estimated statistically. 1961,102 pp.
262 PROCEEDINGS—SPRING JOINT COMPUTER CONFERENCE, 1963