Abstract
The representation of everyday concepts is important for a number of applications, ranging from the Semantic Web to NLP and general AI. We propose here a detailed case study of the Leuven concept database (LCD), which is a rich database of commonsense knowledge, written in natural language. We aim to convert the commonsense knowledge contained in the LCD into a format suitable for implementation and practical application. We then investigate a hybrid approach that combines a syntactic analysis of the surface structure of the LCD entries with a semantic and ontological analysis of those entries, considering also the role of other cognitively-grounded facets of core knowledge. The approach therefore suggests a systematic portfolio of disambiguation modes with the goal of improving the match between everyday meaning of concepts and formal semantics. Finally, we illustrate the practical usefulness of this approach in a concrete computational implementation for concept combination.