Scaling behavior and text cohesion in Korean texts

Abstract

This study examines whether different types of texts, particularly in Korean, can be distinguished by the scaling exponent and degree of text cohesion. We use the controlled growth process model to incorporate the interaction effect into a power-law distribution and estimate the implied parameter explaining the degree of text cohesiveness in a word distribution. We find that the word distributions of Korean languages differ from English regarding the range of scaling exponents. Additionally, different types of Korean texts display similar scaling exponents regardless of their genre. However, the interaction effect is higher for expert reports than for the benchmark novels. The findings suggest a valid framework for explaining the scaling phenomena of word distribution based on microscale interactions. It also suggests that a viable method exists for inferring text genres based on text cohesion.