In this thesis, a rule-based system which resolves Korean anaphora is designed and implemented with integrated Expert System building tool, GURU.
Anaphora is classified into two types - Intra sentential Anaphora and Inter sentential Anaphora. Intra sentential Anaphoras (Reflexive Pronoun and Equi-NP deleted pie) are resolved by syntactic rules, while Inter sentential Anaphoras (Pronoun and Ellipsis) are resolved by Topic elements represented by fuzzy variable. Topic elements and their Topicalities are obtained from postpositions, word order, and Interesting Values.
Concept of topic is adequate to Anaphora resolution in Korean because there are many hints to find topical elements from surface structure and Korean has the language structure in which topic and subject are both important in sentence formation.