A Formalized Contrastive Description of Lexical Units: The Case of Norwegian Modal Nouns

The background of the research is the desirability of reusable lexical resources supporting natural language processing (NLP). The project aims on exploiting the process of creation of bilingual (and monolingual) dictionaries to create more universal and reusable resources, such that would be usable to dynamicaly create different types of dictionaries on demand (e.g. referential dictionaries for professional users, pocket dictionaries for travellers; both uni- and bi-directional dictionaries, aso.) as wel l as reusable resources for NLP tasks. While tradional printed dictionaries are large and detailed, a lot of information contained in them is implicit, cannot easily be digested by computers, and does not reflect all the contextual information of the word s. The goal of the project is to examine the possibilities of a highly structured and formalized (explicit) lexical description and its possible utility as a resource in NLP. One of the important and innovative aspects of the project is a strong emphasis on the role of language variability and stylistic information, which is often beeing underestimated in dictionaries and mostly completely ignored by most NLP projects, however it is essential for parsing and comprehension of more complex texts. Norwegian is used as a suitable language for the proof of concept, because of its wide orthographic variability and existence of two different norms that are not entirely distinct in practice, for instance within the spoken dialects and in some literary works. Nou ns expressing modal relations have been chosen for testing the system, as they are a relatively compact group of words with non-trivial (but not overly complex) syntactic and semantic behaviour and variability, spread from the core to the periphery of the language system.