An Efficient Syntactic Analyzer Using Semantic Connection Units

Abstract

We implement a Korean syntactic analyzer which decreases many ambiguities in syntax parse trees using segmentation and semantic connection units. We use dependency grammar for parsing. Our syntactic analysis system generates all parse trees of a given sentence. So, the number of parse trees of syntactic analysis is many. To decrease the number of parse trees, we suggested semantic connection unit of segmentation. Semantic connection unit is a combined node which consists of all morphemes within a segment. During parsing, our system generates all possible partial parse trees in a segment. There are many partial parse trees after parsing of a segment. We can select some semantic connection units. By using this semantic connection unit, syntactic analysis system can decrease the number of ambiguities of parse trees in Korean syntactic analysis. And also we suggested the pattern rules of predicate to decrease the number of parse trees. There are two types of patterns based on predicate of modifier form, and we found out pattern rules of that predicate pattern. By using this pattern rules, we can decrease the number of ambiguities of parse trees also.