Dependency grammar is used in Korean language parsing for partial free order and omissions of natural language. Korean dependency grammar is defined by the only dependent relation and there are many probabilistic models for dependency grammar.
Probabilistic models solve many problems of parsing, but it cannot resolve some ambiguities because those models ignore grammatical properties of different words. In Korean dependency grammar, sentence connection and verb complement are explained by same probabilistic model.
This thesis proposes a method of using verb caseframe for resolving an ambiguity of dependency between case units and verb. And it gives case-shift patterns and surface caseframe information.
Parsing system consists of two parts, sentence splitting and dependency parser for simple sentences. In this thesis, only dependency parser for simple sentences is constructed.
Simple sentences can be parsed with little information. So paring can be performed more effective process. The proposed model uses caseframe as condition of restrictions on connection, and resolves ambiguities between verbs and case units. Sentence splitting remains further research topic.
