Bookstein shows that information on usage counts and linkages between patrons and items can be retained and utilized while maintaining confidentiality. If one links the patron ID file with the historical use data by using a transforming function which computes easily from patron to circulation data, but not easily from use data to patron identity, some degree of security is maintained.
A retrieval system based on circulation data would be required to return those other items checked out by the patrons who had checked out some percentage of the items on a query list. Both direct matching and a weighted vector model are analyzed to show that efficiency and patron privacy can both be maintained.
A hypertext link may be organizational in the sense that its purpose is to indicate the structure of the document as organized for presentation. Baron, Tague-Sutcliffe, Kinnucan, and Carey test three classes of other than organizational links. Semantic links represent associations among words. Rhetorical links are used to create a path which will convince or educate a reader. Pragmatic links express a relation between a text and its possible use.
Subjects in a browsing task were tested on content assimilation, and surveyed as to attitude toward the system. In the querying task subjects were asked a series of data specific questions using the system while accuracy, time and number of screens viewed were recorded.
Labeling reduced the use of semantic links but had little effect on the use of rhetoric or pragmatic links. No significant learning change was indicated by the use of labels in the browsing task. Labels lead to significantly higher scores in the querying task. Attitudes did not vary in the browsing task, and overall impressions did not vary in the querying task. However, participants without labels found it more difficult to move between topics, and those with labels felt more comfortable moving sequentially through the file than did those without, even though they did not use this technique more than others.
Past work on stemming algorithms has concentrated on the removal of suffixes. Ahmad, Yusoff, and Sembok point out that in languages like Malay or Arabic variant formation depends on prefixes as well as suffixes. Oth-