KNOWLEDGE REPRESENTATION FOR NATURAL-LANGUAGE UNDERSTANDING - THE L(LILOG) APPROACH

Knowledge-based natural language understanding requires the representation of various types of knowledge like linguistic knowledge, conceptual knowledge, taxonomic information, or knowledge about real world objects. Within LILOG, a natural language understanding project for German, the logic-based knowledge representation language L(LILOG) is being used to represent both the semantic background knowledge as well as the information extracted from German texts. L(LILOG) integrates frame-like feature-value descriptions used in the area of computational linguistics into an order-sorted predicate logic framework. The basic design principles of L(LILOG) is presented, give examples how it is used to model real world knowledge, describe the implementation of the first LILOG prototype, and provide a formal semantics definition.