Search results

The application of the variety‐generation technique to the construction of truncated author‐title search keys for data bases of monograph records is described. Instead of the usual fixed‐length keys (e.g. three characters of the author's surname, and the first three filing characters of the title) the method uses strings of characters which vary in length according to the statistical characteristics of author names and titles in BNB MARC files. The number of these strings is variable within wide limits. By choice of appropriate sets of keys, bibliographic search codes with a higher resolving power than the 3,3 key can be produced. The codes also show favourable distributions for direct construction of hash addresses.

A study has been made of the effect of controlled variations in indexing vocabulary size on retrieval performance using the Cranfield 200 and 1400 test collections. The vocabularies considered are sets of variable‐length character strings chosen from the fronts of document and query terms so as to occur with approximate equifrequency. Sets containing between 120 and 720 members were tested both using an application of the Cluster Hypothesis and in a series of linear associative retrieval experiments. The effectiveness of the smaller sets is low but the larger ones exhibit retrieval characteristics comparable to those of words.

Variety Generation involves the selection of sets of character strings, or symbols, which are intended to occur with equal probabilities in bodies of text or sets of text units from a particular source. It is important that the sample used to generate the symbol set should be representative of the data with which the set will be used. An assessment is given here of the amount of variation in symbol sets generated from files of titles and author names from BNB MARC data over a five year period, and a comparison is made with LC MARC. Some of the BNB symbol sets are compared directly, and equifrequency statistics are obtained for the assignment of each symbol set to each file. The differences between the equifrequency statistics are examined by means of an analysis of variance technique.

The use of variety generation techniques in the production of author‐title search codes for files of monograph records is compared with methods based on division hashing. The latter perform better, and evidence is presented to suggest that the reason for this is the lack of statistical independence between the assignments of variety generation symbols to different parts of a field in the record.

The application of a variable to fixed‐length compression coding technique to two bibliographic data bases (MARC and INSPEC) is described. By appropriate transformation of characters or digrams into bit patterns reflecting more accurately the distributions of characters in the data bases, and application of the encoding process, varying degrees of compression can be obtained.

The use of variety generation techniques in the production of fixed‐length degenerate representations for search terms is compared with methods based on division‐hashing. For files of words taken from INSPEC data, the latter perform better, almost certainly because of dependence between assignments of symbol sets. Attempts to overcome the problem produced no real improvement, possibly because the decrease in dependence was accompanied by an increased skewness in the distribution of symbols.

Index entries from the subject indexes to CA can, in general, be converted to ‘normal’ or title‐like phrases by applying simple tests to the positions of prepositions and conjunctions in the entries. Other, more complex, entries can be transformed after somewhat deeper analysis. These manipulations are a necessary preliminary step to the use of the subject‐index language in retrieval. A scheme is outlined for automatically compiling and editing subject indexes by transforming descriptive phrases with regular structure and vocabulary. These transformations, based on the formal structure of language, are shown to be admirably suited to computer manipulation.

Ayers’ recent suggestions for a Universal Standard Book Number, logically generated from a catalogue entry, and therefore applicable restrospectively to bibliographic files, have been implemented and tested on two one‐year cumulations of BNB MARC files. The proportion of unique entries provided by the USBN was found to be about 91%. Revisions to the coding tables were made on the basis of a detailed analysis of the results and of determinations of the frequencies of characters in the data elements used. These resulted in improvements to the method, giving an increase in the proportion of unique entries to approximately 96%.

Computer‐aided systems for all four of the conventional divisions of the field of library housekeeping are now operational in libraries. In Great Britain the only activity not yet operational is that of periodicals control, although two libraries have already used a computer to produce lists of periodicals holdings. The aim of this paper is to describe simply how computer‐aided housekeeping systems work, from the point of view of the librarian, and to mention briefly some typical applications of each type of system.

This is the first annual cumulation of a new indexing service attempting world coverage in geography for periodicals, serials, conference proceedings, and reports. The Institut in Bad Godesberg working with the Zentralstelle für maschinelle Dokumentation (ZMD) records references on a tape typewriter and publishes two‐monthly lists which give the following elements for each entry; UDC number, a set of descriptors, author (both personal and corporate), title, scource, and type of publication.