The snappy title is a bit misleading to the reader who expects to see a report on an existing automatic hypertext-generation system. Certainly steps in formal representation are presented and illustrated by a real document collection, but no resulting system is demonstrated. The authors define, and furnish tutorial illustrations of, a context-free grammar. While a terminal symbol is a terminal symbol is a terminal symbol, an illustration using, say, “Message for:” would be preferable to one using “Vastaanottaja:”; indeed, the authors say, “If the reader of the message does not understand Finnish, she or he will not understand much….” The section on grammar-based models concludes with the introduction of constrained grammars, and is followed by a concrete example of the Canadian Patent Reporter as a source text, including constrained schema from this source.
The authors next define indexing of sources as grammar transformations in general, introducing productions that “hide” structural details and illustrating production rules for the Canadian Patent Reporter. The indexing process is not illustrated, but indexing is described as a hypergraph (consisting of a set of nodes and a set of nonempty subsets of the first set such that all nodes in the first set are contained in the union of all subsets). A subset with two elements is called a link, one with an arbitrary (positive) number of elements an edge, and an ordered subset is called a chain. From this, it is possible to define conventional indexing, relationships, and retrieval operations. Structurally-determined relationships also are included, perhaps derived from the GSML. Navigation and browsing on these bases can be defined.
The section on implementation is disappointing. It follows the definition of the models, operations, and illustrations from a real database, by which time the reader has been led to a point where it is legitimate to ask, “Show me how this works,” rather than “Are there any errors in the model, elements, concepts, or derivation?” Unfortunately, there is no illustration of a system operating on text from the Canadian Patent Reporter or from any other text collection.