We present two methods for evaluating automatically generated hypertex
t links. The first method is based on correlations between shortest pa
ths in the hypertext structure and a semantic similarity measure. Expe
rimental results with the first method show the degree to which the hy
pertext conversion process approximates semantic similarity. The seman
tic measure is in turn only an approximation of a user's internal mode
l of the corpus. Therefore we propose a second evaluation method based
on measuring user's performance using hypertext. Finally, we discuss
the advantages and disadvantages of computer versus human evaluation,
respectively. (C) 1997 Elsevier Science Ltd.