Automatically transforming regularly structured linear documents into Hypertext

s from the year’s technical reports.7 A sample abstract may be seen in printed form as Figure 3. This technical report is issued jointly by UMIACS and by the Computer Science Department; hence it carries two identifying numbers. The abstract was described by the troff source of Figure 4, and the corresponding Hyperties article is shown in Figure 5. Here, the identifying numbers are used as the article’s title, and the article’s body is formed by rearranging the components of the original listing. The abstract’s title is also used as the Hyperties article’s short description.8 The article’s body is augmented with links to a global index and to indices for each of the issuing departments, as well as links directly UMIACS–TR–88–4 CS–TR–1973 Title: Efficient Stochastic Gradient Learning Algorithm for Neural Network Author: Y.C. Lee, Department of Physics and Astronomy Efficient first and second order adaptive learning algorithms of stochastic gradient descent variety are described. Both algorithms can automatically adjust step sizes to achieve optimum convergence rates. Various theorems concerning the convergence properties of these algorithms are discussed. Figure 3. A sample UMIACS abstract .(l L .b UMIACS\-TR\-88\-4 .b CS\-TR\-1973 .sp Title: Efficient Stochastic Gradient Learning Algorithm for Neural Network Author: Y.C. Lee, Department of Physics and Astronomy .)l .sp .pp Efficient first and second order adaptive learning algorithms of stochastic gradient descent variety are described. Both algorithms can automatically adjust step sizes to achieve optimum convergence rates. Various theorems concerning the convergence properties of these algorithms are discussed. .sp Figure 4. troff markup of sample UMIACS abstract 7 The 1988 UMIACS abstract listing contained 102 abstracts. The troff-based input markup file was approximately 125 000 bytes in length. 8 Consequently, the abstract’s title is used both in the Hyperties article body and also in the short description.