Special Librarian - Master of, or Slave to the Machine?


Eugene Garfield, Ph.D., President
Institute for Scientific Information
325 Chestnut Street
Philadelphia, Pa. 19106

Paper presented at the
57th Special Libraries Association Annual Convention
Minneapolis, Minnesota
May 31, 1966


The situation of the special librarian today, vis-a-vis the computer, is understandably characterized by semantic confusion. The purveyors of machine systems have frequently clouded the issue by "gee-whiz" allusions to the capabilities of electronic computers. Frustration, shyness, or fear have commonly and inevitably accompanied the confusion. On the one hand, librarians have been told to expect the impossible from computers. On the other hand, they have feared the worst. Some have developed an unnecessary apprehension that computers will displace librarians. Of course, there are some librarians with professional degrees who are really "clerks-in-camouflage." But even to these people, I can offer some solace. Automation in library systems has created a need for new kinds of bibliographically trained clerks. For example, the professional and clerical personnel employed by ISI have, through contact with our computer systems, often become better trained bibliographically than are their counterparts in traditional systems. And their coordinated efforts augment and vastly improve the mechanized outputs of the computer.

When I first planned this talk last year, it was going to be a brief description of ISI's services of special interest to librarians (1). However, after a conversation with Mr. Aspnes in March, my talk evolved into a comparison of at least four leading machine systems -- ISI, ACS, NLM and NASA -- and what these systems required of the special librarian. More recently, however, it became clear that the preferred topic for discussion is the one implied by my newly chosen title. Mr. Aspnes refers to this subject as "the special librarianís responsibilities in communicating with the machine," There can be no doubt that the librarian can play an important role in machine systems. Whether he chooses to do so will determine whether he will be the slave to or master of the machine. In the latter case, the librarian must be able to communicate with the machine.

Many magnetic tape systems are available today which purport to serve the individual library client. In all of them, the librarian or the patron is expected to participate in an active way. This participation is essential from the initial construction of the search profile through the delivery of periodic reports and the redesign of profiles in terms of responses and changing interests. In all of the systems, the librarian, either alone or in consultation with the scientist, can help act as a translator. The special librarian must translate the information requirement of the patron - that is, a series of questions expressed in the userís language - into the language of the system employed. Please note that I did not say actual machine language because that is neither true nor necessary. Whether a particular computer is binary coded decimal (bcd) need not concern the librarian or patron. However, you will be required to understand the limitations and logic of the particular search system employed. Sometimes the logic will be rather involved. Other times, the logic will be absurdly simple. In either case, basic grasp of the logic is essential.

In all of the machine systems available today, including ISI's the use of words as "subjects" is involved. However, the natural language of authors, as expressed in titles, is a basis of selection in the ISI (2) and ACS (3) systems. In the NLM (4) and NASA (5) systems, the basis of selection is "unnatural" language. Controlled terms from a thesaurus are selected by indexers in the hope of making the indexing more consistent. Ordinarily, the price the user pays for this benefit is an increase in the number of unwanted documents retrieved.

To reiterate, the librarian must translate the original search request into a list of words, thesaurus terms, or subject headings that will be used in the search system to select those documents of possible interest to the client. In other words, the librarian helps construct a search "profile." Most of you have encountered difficulty in preparing such profiles. If so, let me assure you that there is no need to suffer from any feeling of inferiority. The construction of a search profile is one of the most delicate and sophisticated assignments that can be given to a special librarian. According to the circumstances, you would be well advised to engage in a series of dialogues with your clients to overcome these difficulties which are innate to the research process itself.

Before discussing the reasons for the special difficulties of word profiles, it is important to mention that the ISI searching system also accepts a unique type of search profile - a citation profile. In this case, the librarian does not translate a search request into a list of words, but rather into a list of citations (a bibliography) which describes the clientís field of interest.

Why is the task of preparing a search profile sometimes so painful? The scientist may or may not be able to give you a list of terms that will delineate his field of interest. This list will always suffer from certain inadequacies. These inadequacies are the result of problems common to all translation work. Translation always requires the resolution of ambiguity, synonymy, near-synonymy, etc. One manís adrenaline is another manís epinephrine. One person says content analysis, while another says subject analysis. An American says structural linguistics, while his Soviet counterpart usually says mathematical linguistics. No matter how carefully the client chooses his terms, with or without your help, he will never be able to anticipate every word or word cornbination that might occur which he could interpret as being relevant to his interests. Neither can you. However, if you are acting as his proxy, then you at least should be prepared to say that you know better than he, how to query the information system and what are its limitations. Do not expect the computer to replace the subjective notion of relevance which only you or the user can bring to the system (6).

Scientists may have trouble in providing a list of effective searching terms. Similarly, a scientist may initially find it strange to be asked for a list of publications defining his field. What he will actually be doing is providing a list of those publications in which he has displayed interest in the past. These will be as closely similar to what he hopes to find and can stipulate in advance from memory or from your retrospective search efforts. You will find this is surprisingly effective in anticipating the very thing that may be lost in a word-limited profile.

We have found that the vast majority of scientists using these SDI systems are regularly writing papers or reports. The list of papers cited in a manís most recent paper will frequently serve as an excellent starting bibliography for his citation profile. Incidentally, Mary Stevens of the National Bureau of Standards (7) has reported on some interesting experiments which would confirm this. Similarly, the title words or index terms from this list of earlier papers is a good starting point for a word profile. The combination of both capabilities, word and citation profiles, will satisfy most user needs.

There are many other facets of library systems and automation which could be discussed in considering whether the special librarian is to be master of or slave to the machine. I have emphasized here the librarianís role as an interpreter -- a translator in the search process, whether for selective dissemination or for retrospective search. I have emphasized this function because by definition the special librarian is one who specializes in some branch of knowledge, though we know that he may assume all or part of the administrative functions of the general librarian. I have also tried to present an optimistic picture. Combined with a severe shortage of librarians in general, the need for every librarian to regularly re-educate himself presents marvelous opportunities for individual advancement, service, and professional progress which can, within the next decade, eliminate the well known stereotype librarian. That image will become a relic of history. I donít think any of you will have regrets about this by-product of the impact of technology on our lives.


1. back to text   E. Garfield, "ISI Services in the Design of Small User Systems." Paper presented at the 151st National Meeting of the American Chemical Society for the Division of Chemical Literature Symposium on Problems of Small Information Groups, Pittsburgh, Pa., March 25, 1966.

2. back to text  C.R. Sage, R.R. Anderson and D.R. Fitzwater, "Adaptive Information Dissemination," American Document ation 16(3), 185-200 (1965).

3. back to text C.N. Rice, "A Computer-based Alerting System for Chemical Titles,"Journal of ChemicalDocumentation 5(3), 163-165 (1965).

4. back to text U.S. Public Health Service, MEDLARS Story at the National Library of Medicine, 1963, 74 pp.

5. back to text W.T. Brandhorst and P.F. Eckert, "NASA Search System Analysis Sheet," (letter-to-the-editor) American Documentation 16(2), 124-126 (1965).

6. back to text E. Garfield, "Can Citation Indexing Be Automated?" in M.E. Stevens,
V.E. Giuliano and L.B. Heilprin, Eds., Statistical Association Methods
for Mechanized Document at ion, Symposium Proceedings, Washington, 1964,
(NBS Misc. Pubi. #269, December 15, 1965), pp. 189-192

7. back to text M.E. Stevens and G.H. Urban, "Automatic Indexing Using Cited Titles," in M.E. Stevens, V.E. Giuliano and L.B. Heilprin, Eds., Statistical Association Methods for Mechanized Documentation, Symposium Proceedings , Washington 1964, (NBS Misc. Publ. #269, December 15, 1965), pp. 213ó215.