Search results

With the growth in multiple on‐line search systems providing access to bibliographic data bases, exclusive availability of data bases is quickly disappearing. The searcher equally familiar with more than one system providing access to the same data base needs to apply some objective criteria in choosing a particular search system. Comparing three major search systems, we emphasize system differences in file structure, data fields and search features, and discuss how these differences influence search strategy and results. Results obtained from the CACondensates, ERIC, INFORM and NTIS data bases serve as examples for discussing the differences among the three search systems. Differences in postings among the systems are most often explained in terms of expected differences due to file structures, data fields and search features. Some unexpected differences between the systems are described and are assumed to be attributable to problems encountered during the preprocessing of the tape.

First of all, I must apologise for the interval between this VINE and the last. Unfortunately VINE's production cycle is growing longer as automated library systems become more complex, and consequently more time‐consuming to write up. Moreover, in this issue I have attempted in certain articles, for instance those on COM bureaux and the Telepen, to adopt a thematic approach to the subject, rather than reporting on individual projects. The process of cross‐checking the details of such articles with all the organisations concerned has been partly responsible for the delay in publishing VINE 17. Nevertheless in the long terms I still hope to increase the frequency with which VINE is published, thereby increasing its currency and decreasing the size of each individual issue.

For a number of reasons, I have decided to begin this issue of VINE with an editorial. In the first place, as many of you will already know, at the end of June I gave up my post as Information Officer for Library Automation in order to become an Assistant Librarian in the Applied Sciences Library of University College, Cardiff. Since the Information Officer for Library Automation is ex officio editor of VINE, this means that I will also be handing over responsibility for this publication to my successor, when he or she is appointed. Unfortunately, there will be a hiatus before my successor can take over, so that for a short while, I will continue to edit VINE. However, because of my commitments in Cardiff, I will be unable to provide news of recent developments in U.K. library automation as I formerly did. Thus, in this issue of VINE and the next, I will concentrate on descriptions of automated library systems which I visited whilst on a study tour of North America in April and May of this year. Once my successor takes over, VINE will revert to being a source of up‐to‐date information about automation in specifically U.K. libraries.

Proposals are presented for specifying a computer system for storing and searching library records in a single library or a group of associated libraries. The system provides for on‐line searching at each library and off‐line production of accessions lists, abstracts bulletins and bibliographies, as well as SDI service. The use of skilled effort is shifted from the in‐put operation to the searching stage by the use of selected natural language keys, but searching for known items is made easier than in more traditional methods by extending the use of post‐coordination to catalogue data.

This paper discusses some particularly desirable features of the user‐computer dialogue involved in query formulation in an on‐line information retrieval system. The methods used in the Culham RIOT II system, designed to give on‐line access to a data base of 25, 000–50, 000 references, are described. To facilitate query formulation a matrix technique is employed and the most recent titles may be displayed while the query statement is under construction and before a full search is requested.

The possibility of having access to all the world's literature from a single computer terminal stimulated the imagination of the research workers in the late' sixties. It was this goal and the fascination of the co‐operation between man and machine, that inspired the major changes that have taken place in Information Retrieval over the past ten years.

Evaluations of on‐line information‐retrieval systems have been largely dependent upon monitoring of users' searches and the cooperation of users in interviews and questionnaires. Since users have a variety of information needs and levels of experience, the evaluation process has been difficult. This paper describes and presents the results of a series of experiments designed to evaluate selected features of a full‐text information‐retrieval system in a controlled environment. Features are evaluated on the basis of their development and implementation cost and their effect on user performance and system load. A figure of merit is developed for each feature, and the results of statistical tests to determine the validity of these measurements are presented.

Currently available on‐line systems historically have been used as retrospective search methodologies. With interest in on‐line retrieval increasing by leaps and bounds, the application of these inverted‐file oriented systems to selective dissemination of information (SDI) services bears investigation. Within given user‐specified demands and system constraints, what is the best method of providing an SDI service? A typical batch system is compared with four types of commonly available on‐line SDI systems. The operating parameters of each are outlined, and estimates of bottom‐line costs are given.

Two types of linear on‐line search have been implemented on a data base of 100,000 INSPEC records, and compared with an inverted search program on the same file. It was found that the linear search which does a character‐by‐character scan of the whole file was too slow to be useful as an on‐line search, but that the second linear search, which uses a linear file of keys to select records for full scanning, was useable on‐line. An algorithm suitable for the on‐line searching of the linear file of keys is presented with run time nearly independent of the number of terms in the search query. Neither of the linear searches can compete with an inverted search in terms of user convenience, but they do not require as complex file‐maintenance software, they use less processing time to build indexes, and they have lower storage requirements than inverted files.

VINE is a Very Informal Newsletter produced three times a year by the Information Officer for Library Automation and financed by the British Library Research & Development Department. It is issued free of charge on request to interested librarians, systems staff and library college lecturers. VINE'S objective is to provice an up‐to‐date picture of work being done in U.K. library automation which has not been reported elsewhere.