Might a semantic lexicon support hypertextual authoring?

Might a semantic lexicon support hypertextual authoring? Roberto Basili, Fabrizio Grisoli, Maria Teresa Pazienza Dipartimento di Ingegneria Elettronica, Universita' di Roma, Tor Vergata Via della Ricerca Scientifica, 00133 Roma (ITALY) e-maih { rbas, paz ienza } 8rovvxl. ccd. utovrm, it Abstract It is common opinion that current hypertextual systems do not allow to express objectively the information content of documents, but only the view of the "author". The hyperlink building requires an heavy and highly specialised human intervention: this task is very expensive whenever possible! A different approach, based on NLP methodologies, aiming at automatizing the development of an hypertext, is hereafter proposed. Anchorage points are inferred both from content and structure of documents. A semantic lexicon based on conceptual graph structures is used to guide text understanding. Contextual roles are introduced to model domain specific concepts relevant to the navigation. An off-line activation of useful links has been defined according to explicit user specifications. A simple declarative language (HyDeL) for the definition of such links is available to the user to create his own views on the document base. HERMES is a prototype system implementing our approach. a The paper discusses the semantic processing of document base and highlights the performance of different hypertextual systems derived by HERMES over different languages and knowledge domains. Introduction An intuitive representation of an hypertext is depicted as a directed graph, in which nodes are documents, and arcs are links. Browsing an hypertext means to move from a document (source node) to another (destination node) through a link. Specific words, named anchors, are considered more important than others, as they are source or destination of a link. Many different criteria have been used to describe existing hypertextual systems. According to (Frisse, Cousin 92) three different "levels" in modelling an hypertext should be underlined: physical, logical and browsing semantics levels. At physical level, the model addresses the data organization on the storage devices. At logical level, relationships among data are described. Finally, the browsing semantics oriented models are used to describe human interaction modalities. Main actors in hypertext systems are the author and the user: the author organizes the overall hypertextual structure defining the number and type of links between documents, thus imposing a particular navigation modality in the document base 1. The user may browse the hypertext, following only the defined hypertextual structure. Authoring is a major problem in hypertextual systems: the requested human effort explodes with the document base size. An incremental approach is often impossible either because the insertion of a new document requires a reprocessing of the whole document base, or the new document may introduce new perspectives. The management of very large document bases would suitably profit of an automatic approach to authoring. Furthermore it is simply unrealistic that the author knows what is of interest in any document, at any time. Manual authoring thus bases on a static source of information modelled once; it should be desirable a more objective and exhaustive processing. Our idea of hypertext is …

Content truncated to 500 words