A Multi-view Hyperlexicon Resource for Speech and Language System Development

New generations of integrated multimodal speech and language systems with dictation, readback or talking face facilities require multiple sources of lexical information for development and evaluation. Recent developments in hyperlexicon development offer new perspectives for the development of such resources which are at the same time practically useful, computationally feasible, and theoretically well– founded. We describe the specification, three–level lexical document design principles, and implementation of a MARTIF document structure and several presentation structures for a terminological lexicon, including both on demand access and full hypertext lexicon compilation. The underlying resource is a relational lexical database with SQL querying and access via a CGI internet interface. This resource is mapped on to the hypergraph structure which defines the macrostructure of the hyperlexicon.