Modelling spatial knowledge on a linguistic basis. Theory, prototype, integration (Q1202142)

The book develops a theory about knowledge of spatial objects which integrates aspects from cognitive linguistics and artificial intelligence. Based on this theory, a prototype system, OSKAR, has been implemented in Prolog. First, the major elements of the linguistic theory of spatial knowledge and spatial reasoning are worked out in detail. The semantics of natural language object names and dimensional expressions is analyzed on the basis of \textit{Lang} and \textit{Bierwisch's} two-level model which relates the level of language-specific word meanings (semantic form expressed by grammatically coded dimension assignment parameters) to the level of language-independent conceptual knowledge (conceptual structure expressed in terms of object schemata). The linguistic analysis and formal account concentrates on a relevant subset of spatial expressions, viz. dimensional terms. Next, a detailed overview of the design and knowledge representation capacity of the OSKAR prototype system is given including fragments of Prolog code. OSKAR is capable of understanding, i.e. reasoning about German spatial expressions. In particular, it accounts for the dimensional designation of objects, categorization of spatial objects,and positional specification of objects. The final chapter sketches OSKAR's integration into the knowledge representation component of the LILOG natural language understanding system.