This thesis describes the design and implementation of a Korean generation system using the semantic-head-driven method. This system improves portability and expandability compared to the previous Korean generation systems.
To enhance the portability, the proposed system uses a general tree-formed semantic representation so that this generation system can be easily used by many other application systems: machine translation systems, dialogue systems, natural language database query systems, etc. To ensure expandability, the system is divided into a syntactic generator and a morphological generator. The syntactic generator consists of the rule part and the generation engine. Each rule can be expressed as an augmented CFG form to overcome the shortcomings of the semantic-head-driven method and to increase the readability.
The generation system can generate various types of Korean sentences that include embedded, conjunctive, passive sentences and so on. Several morphological phenomina such as conjugation, vowel harmony, modality and post-ending are also taken into account.