An experiment showing that 3, 1, 1, 1 search keys derived from titles are sufficiently specific to be an efficient computerized, interactive index to a file of 135,938 MARC II records.This paper reports the findings of an experiment undertaken to design a title index to entries in the Ohio College Library Center's on-line shared cataloging system. Several large libraries participating in the center requested a title index because experience in those libraries had shown that the staff could locate entries in files more readily by title than by author and title. Users of large author-title catalogs have long been aware of great difficulties in finding entries in such catalogs. Since the center's computer program for producing an author-title index could be readily adapted to produce a title index, it was decided to add title access to the system.A previous paper has shown that truncated three-letter search keys derived from the first two words of a title are less specific than authortitle keys ( 1). Earlier work had revealed that addition of only the first letter of another word in a title improved specificity ( 2) . Therefore, the experiment was designed to test the specificity of keys consisting of the first three characters of the first non-English-article word of the title plus the first letter of a variable number of consecutive words.The experiment was also designed to produce an index that catalogers could use efficiently and that would operate efficiently in the computer system. It was assumed that the terminal user would have in hand the volume for which an entry was to be sought in the on-line catalog. The index was not to be designed for use by library users; subsequent experiments will be done to design an index for nonlibrarian users.Other investigations into computerized, derived-key title indexes include

<p class="p1">Development and <span class="s1">implementation of </span>an off-line <span class="s1">catalog card </span>production <span class="s1">system </span>and <span class="s1">an </span>on-lin<span class="s2">e </span><span class="s1">shared cataloging </span>syst<span class="s2">e</span>m are described. In off-lin<span class="s2">e </span>production<span class="s2">, </span>average <span class="s1">cost </span>p<span class="s2">e</span>r <span class="s1">card </span>for 529<span class="s2">,8</span>93 <span class="s1">catalog </span>cards <span class="s1">in </span>finished form and alphabetized for filing was 6.57 cents. An account is <span class="s1">given </span>of <span class="s1">system </span>design <span class="s1">and equipment </span>selection for the on-line system<span class="s3">. </span>File organization <span class="s1">and </span>pro<span class="s2">gra</span>ms <span class="s1">are </span>described, <span class="s1">and </span>the on-line <span class="s1">cataloging system </span>is discussed. The system <span class="s4">is </span><span class="s2">eas</span><span class="s4">y to use</span><span class="s2">, </span>efficient, reliable, <span class="s4">and </span>cost <span class="s4">beneficial.</span></p>

<p class="p1">An experiment was conducted to design a corporate author index to a large bibliographic file. The nature of corporate entries necessitates a different search key construction from that of personal names or <span class="s1">titles. </span>Derivation of a search key to <span class="s1">select </span>distinct corporate entry records <span class="s2">is </span>discussed<span class="s3">.</span></p>

<p class="p1"><span class="s1">A </span>test of validity of earlier <span class="s2">findings </span>on <span class="s2">3</span><span class="s3">,3 se</span><span class="s2">arch-k</span><span class="s3">e </span><span class="s2">y r</span><span class="s3">e</span><span class="s2">tri</span><span class="s3">ev</span><span class="s2">al </span>from an in-proc<span class="s3">ess </span>file for r<span class="s3">e</span>trieval from <span class="s4">a </span>MARC fil<span class="s3">e. </span>Probability of numb<span class="s3">e</span>r <span class="s4">of </span><span class="s3">en</span>tri<span class="s3">es </span>r<span class="s3">e</span>tri<span class="s3">eve</span>d p<span class="s3">e</span>r reply <span class="s5">is </span><span class="s3">esse</span>ntially th<span class="s3">e </span><span class="s4">same </span>for both <span class="s4">files.</span></p>

No abstract