KNOWLEDGE REPRESENTATION FOR NATURAL LANGUAGE UNDERSTANDING

Determining the meaning of a sentence in a natural language requires an efficient search mechanism to find the corresponding point in a very large space of sentence meanings. Examination of large search spaces must rely on constraints to guide the search process and provide satisfactory performance. However, existing knowledge representation techniques used in AI, both classical and connectionist, do not satisfy all the requirements needed to represent adequately the semantics, whereby neither of these techniques meet satisfactorily the most important requirement proper context representation. We propose a knowledge representation technique named Hierarchical Semantic Form that can be used, together with the Space Of Universal Links (SOUL) algorithm, to represent the semantics adequately. To check the viability of the proposed solution we have implemented a prototype Semantic Web service that provides information about flight timetables, defined in a natural language within an ordinary HTML page, using natural language queries.