Danish knowledge patterns and word sketches for semi-automatic extraction of terminological information

Abstract

Many articles on terminological information extraction using knowledge patterns are based on studies of corpora in languages for which extraction is supported by grammars. Languages lacking free access to advanced grammatical resources receive less attention. This paper evaluates the results of semi-automatic extraction of Danish knowledge-rich contexts comprising type-of relations. The knowledge patterns used for the extraction are based on the valency of Danish verbs. This pragmatic character-string approach is tested using the query system Sketch Engine. A Danish corpus on electric cars is used, the aim being to create a generic concept system, and assessing the degree of completeness that may be attained for such a concept system. Then, I shall investigate the influence on terminological workflow exerted by various approaches differing in respect of the stage of development of the resources and the tools available. Also, workflow strategies will be proposed.