In chart parsing, a lot of time is wasted to look for edges which require the relevant category in order to apply fundamental rule of a chart parser. Redundant edges degrade the efficiency of the parser greatly. Reducing these edges increases the efficiency of the parser and also removes ambiguous syntactic interpretations. This thesis presents a method to decrease these edges by using syntactic interpretations. This thesis presents a method to decrease these edges by using syntactic relations. A syntactic relation is a dependency relation between two syntactic constituents according to their functions and types. As a result of this study we found that, in Korean, syntactic relations turned out to be very effective constraints in pruning the edges of no value. To determine syntactic relations, we develop a syntactic-relation table for all possible types and functions of constituents in Korean sentences. By using this table, the information of dependencies in input sentence can be determined. After determining dependency relations in a sentence the parser uses them for reducing implausible edges.
The empirical result shows the edge number is decreased in the ratio of 20~70% about the used sentences for the experiment. The system has been implementaed in "C" programming language and LISP on the UNIX. The grammar is written in ACFG (Augmented Context Free Grammar).
