한국어 형용사 유의어의 군집 구조와 변별 요인 분석: 행동 프로파일 분석을 중심으로
Cluster Structure and Distinctive Factors of Korean Adjectival Synonyms: A Behavioral Profile Analysis
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study aims to identify the distinctive usage patterns of seven Korean adjectival synonyms that express the meaning of degree of intensity. For this purpose, a corpus-based Behavioral Profile (BP) analysis methodology was applied. A total of 1,400 corpus instances were annotated based on 12 morpho-syntactic and semantic variables (ID tags). Hierarchical Agglomerative Cluster (HAC) analysis and statistical validity tests (AU/BP values, Silhouette and Dunn Index) were then used to examine the hierarchical structure of the synonym set. Cramer’s V and effect sizes were further analyzed to determine the key discriminative factors driving the clustering process. The analysis determined that the optimal number of clusters was four: {Gwahada, Neomuhada}, {Geuksimhada, Simhada}, {Gwadohada, Jinachida}, and {Gwabunhada}. In the first clustering stage, morpho-syntactic variables—such as pre-final endings, sentence structure, and predicate structure—together with the semantic variable of semantic preference (formal vs. substantive meaning) emerged as primary discriminators. From the second stage onward, semantic variables had greater influence than morpho-syntactic ones. Specifically, semantic prosody (negative vs positive, neutral) was the major discriminator in Stage 2, while semantic preference (entity & situation-oriented vs. action & property-oriented meanings) characterized Stage 3.
Abstract
This study aims to identify the distinctive usage patterns of seven Korean adjectival synonyms that express the meaning of degree of intensity. For this purpose, a corpus-based Behavioral Profile (BP) analysis methodology was applied. A total of 1,400 corpus instances were annotated based on 12 morpho-syntactic and semantic variables (ID tags). Hierarchical Agglomerative Cluster (HAC) analysis and statistical validity tests (AU/BP values, Silhouette and Dunn Index) were then used to examine the hierarchical structure of the synonym set. Cramer’s V and effect sizes were further analyzed to determine the key discriminative factors driving the clustering process. The analysis determined that the optimal number of clusters was four: {Gwahada, Neomuhada}, {Geuksimhada, Simhada}, {Gwadohada, Jinachida}, and {Gwabunhada}. In the first clustering stage, morpho-syntactic variables—such as pre-final endings, sentence structure, and predicate structure—together with the semantic variable of semantic preference (formal vs. substantive meaning) emerged as primary discriminators. From the second stage onward, semantic variables had greater influence than morpho-syntactic ones. Specifically, semantic prosody (negative vs positive, neutral) was the major discriminator in Stage 2, while semantic preference (entity & situation-oriented vs. action & property-oriented meanings) characterized Stage 3.