Janghwan Kim, Woosung Jang, R. Young Chul Kim, Journal of Internet Computing and Services, Vol. 26, No. 6, pp. 75-82, Dec. 2025
10.7472/jksii.2025.26.6.75, Full Text:  HTML
Keywords: Korean Natural Language Requirement Specifications, Textual Analysis, C3Tree, Natural Language Processing

Abstract

Recently, there has been a rapid growth in large language models (LLMs), sophisticated AI systems that can understand, generate, and manipulate human language. Most AI experts mention prompt engineering, RAG, or fine-tuning as key factors for achieving success with LLM. Even in the field of requirement engineering, we also handle requirement specifications written in the Korean language. However, this language has complex grammatical structures and diverse morpheme combinations, which are crucial factors affecting the performance and interpretation of natural language processing (NLP) models. We attempt to analyze the unique and complex grammatical structures of the Korean language, including complicated, compounded, and simple sentences. As requirement engineers, we should make it challenging for AI machines to understand and process the sentences they learn accurately. To solve this issue, we propose a simple sentence generation mechanism with the C3Tree model. By converting complex sentences into simpler forms, our approach aims to reduce sentence redundancy and also enhance the quality of training sentences. Then, we try to learn the generated simple, compounded, and complicated sentences using AI Machine Learning models to enhance the F1-Score. As a result, we expect to enhance the accuracy of Korean NLP models and improve data quality, thereby making them more effective for various AI applications.

Statistics
Cite this article