No abstract

No abstract

As an introduction to the papers in this special issue, some of the major problems facing investigators evaluating information retrieval systems are presented. These problems include the question of the necessity of using real users, as opposed to subject experts, in making relevance judgements, the possibility of evaluating individual components of the retrieval process, rather than the process as a whole, the kinds of aggregation that are appropriate for the measures used in evaluating information retrieval systems, the value of an analytic or simulatory, as opposed to an experimental, approach in evaluating retrieval systems, the difficulties in evaluating interactive systems, and the kinds of generalization which are possible from information retrieval tests. © 1996 John Wiley & Sons, Inc.

No abstract

A model is presented for converting a collection of documents to hypertext by means of indexing. The documents are assumed to be semistructured, i.e., their text is a hierarchy of parts, and some of the parts consist of natural language. The model is intended as a framework for specifying hypertextual reading capabilities for specific application areas and for developing new automated tools for the conversion of semistructured text to hypertext. In the model, two well-known paradigms—formal grammars and document indexing—are combined. The structure of the source text is defined by a schema that is a constrained context-free grammar. The hierarchic structure of the source may thus be modeled by a parse tree for the grammar. The effect of indexing is described by grammar transformations. The new grammar, called an indexing schema, is associated with a new parse tree where some text parts are index elements. The indexing schema may hide some parts of the original documents or the structure of some parts. For information retrieval, parts of the indexed text are considered to be nodes of a hypergraph. In the hypergraph-based information access, the navigation capabilities of the hypertext systems are combined with the querying capabilities of information retrieval systems.