본문 바로가기
  • Home

Automated Scoring of the Content and Task Fulfillment Domain in L2 Korean Writing Assessment

  • Korean Language & Literature
  • 2026, (133), pp.495~516
  • Publisher : Korean Language & Literature
  • Research Area : Humanities > Korean Language and Literature
  • Received : June 19, 2026
  • Accepted : July 20, 2026
  • Published : July 31, 2026

Baek, Jae-pa 1

1동아대학교

Accredited

ABSTRACT

The purpose of this study is to develop and validate an automated scoring model for the Content and Task Fulfillment domain in L2 Korean writing assessment. To this end, a dataset consisting of 149 learner essays on the topic Expansion of Surveillance Camera Installation was constructed from the Korean Learner Corpus developed by the National Institute of Korean Language. Character count, S-BERT-based document similarity, and topic entropy-based semantic richness were used as predictive features, and automated scoring models were developed using Random Forest and Logistic Regression algorithms. The results showed that both automated scoring models achieved high predictive performance, with AUC values ranging from 0.8952 to 0.9001, an accuracy of 0.7667, and F1-scores ranging from 0.7327 to 0.7598, demonstrating their effectiveness in predicting expert-assigned scores. Feature importance analysis revealed that character count was the most influential feature, while document similarity and semantic richness, both semantic features, also contributed to score prediction. This study contributes to the field by proposing and validating scoring features applicable to automated scoring of the Content and Task Fulfillment domain, thereby providing a foundation for future research on automated scoring of L2 Korean writing.

Citation status

* References for papers published after 2025 are currently being built.