Date of Award
Doctor of Philosophy
The efficient design and maintenance of information retrieval systems are still important considerations, even with the current availability of faster computer hardware and cheaper secondary storage. Informetrics, the quantitative study of information and its use, could provide the systems designer and analyst with tools with which to aid in decision making for optimising system performance.;This study examines how informetrics can be used to help the systems designer in deciding what types of file structures would provide the best performance for a given type of information system environment. It is hypothesized that for varying index term distributions and system term usage patterns, different file structures provide more timely access and require less space.;A factorial design simulation study was carried out to examine the retrieval performance and space requirements of several types of hypothetical information retrieval systems. Performance estimates were based on simulation runs varying system parameters which included the index term distribution, the term selection relationship and number of index terms. Results under different file structures were compared using multi-way Analysis of Variance.;It was found that different structures were better suited for different retrieval environments. The chained hashing structure, in most cases, provided the best retrieval performance and most economical space requirements where gradually decreasing term distributions were present. However, the proposed modified hashing structure performed better in situations where steep term distributions existed. Two variations of the BIM-tree (Balanced Implicit Multiway tree) also provided good performance where steep term distributions were encountered and where the term selection relationship favoured the retrieval of terms with larger postings lists.
Wolfram, Dietmar, "The Application Of Informetrics To Information Retrieval Systems Design: A Simulation Study" (1990). Digitized Theses. 1978.