Constructing Knowledge Spaces From Linguistic Resources

Language documentation aims at the creation of a representative and long last-ing, multipurpose record of natural languages [1]. It contributes to the mainte-nance, consolidation and revitalizing of endangered languages and thus safe-guards the full range of their uses. Such language documentation also contri-butes to the description of cultural practices of a speech community. Our aim is to enrich this cultural documentation by allowing users to link linguistic infor-mation of lexica and annotated media recordings with ontological information in a multimedia web-based lexicon tool. Our approach is centered around the crea-tion of knowledge spaces (KS), where users model a world of concepts and their interrelations for which the organisation of lexical and cultural data is based on categorisation patterns made by the speech community members. Resources. The DOBES archive for endangered languages hosts a rich set of primary resources (audio and video recordings) and annotations for about 35 languages [2]. For Marquesan and Tuamotuan trilingual lexica have been cre-ated comprising approximately 3000 and 850 lexical entries each. A lexical en