THEORY OF SEARCH KEYS AND APPLICATIONS IN RETRIEVAL TECHNIQUES USED BY CATALOGERS

This paper constructs a model for studying the performance of search keys of several types (such as, e.g., author/title keys of the form 4/4, 3/3, 3/1/1/1, and so on), and gives a criterion for deciding whether or not to use one (or several) more slashes (/) (i.e., more truncated title words) in a certain system (e.g., an automated library catalog). Some mathematical theorems on search keys are proved, given the probability distribution of letters (more general: symbols) in words. We then study the effect (on search key performance) of enlarging the "alphabet," i.e., when adding new symbols, that can be used in forming the search keys. Changing the alphabet might cause a change of optimal search keys. Criteria for this (not) to happen are given. The last part of the paper deals with the difference in average performance (averaged over all possible systems) of search keys with less or more slashes (/). In general, we can prove that, on the average, introducing slashes does not improve the retrieval performance.