Searching Versus Browsing in Hypertext

Publisher Website
Google Scholar
Abstract
Our goal is to elaborate the relationships among tasks, user types, and tools for hypertext. We built a computer system for automatically converting books into hypertext and then converted one book into Emacs-Info, Guide, HyperTies, and a variant of SuperBook, called MaxiBook. Three classes of users were studied: experts, novices, and trainees. Users performed two question-answering tasks: one for a search question and one for a browse question. The answer to a search question could be found in one section of the book, whereas the answer to a browse question occurred across sections. Answers were evaluated for quality and speed. Experts did good searching with Guide and MaxiBooks, but the best browsing with paper. For novices, paper was the best tool on all counts — for quality and speed, for searching and browsing. Trainee experiences were similar to those of novices. Based on our results and by contrasting models of the various tools, we infer that a good hypertext system should simulate the ‘index’ and ‘arbitrary jump’ of paper and the ‘dynamic outline’ and ‘implicit index’ of MaxiBook. While sophisticated features of our hypertext, such as an alternate outline, did not prove helpful, we show that the class of ‘understanding tasks’ is much larger than that of searching or browsing tasks and speculate that alternate outlines support some ‘understanding tasks’.