Missing link? Comparison of manual and automated linking in hypertext engineering

Abstract

Most discussions of hypertext usability are not founded in empirical measurement but more on conjectures based on personal experience. In this paper we report on two empirical analyses of hypertext usability, focusing on the quality of links produced by different means. We conducted two experiments to test the predicted relevance and the evaluated relevance of links, that is, where links are evaluated either before or after they are traversed. In order to evaluate these two kinds of relevance, we conducted two experiments where a hypertext document was created from a printed text. In each experiment we compared the relevances of three different sets of links. One set was created by a human author, whereas the second set was created automatically using the HEFTI (Hypertext Extraction From Text Incrementally) model for converting text into hypertext. We also generated a third set of links by assigning links randomly between nodes. The main goal of this research was to develop empirical tests that evaluate the usability of hypertext links. A second goal was to test the validity of automatically generated links using the HEFTI model. In this paper we detail the two experiments, and discuss their implication for methods, of hypertext usability assessment and design.