Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UKTA: Unified Korean Text Analyzer

Seokho Ahn, Junhyung Park, Ganghee Go, Chulhui Kim, Jiho Jung, Myung Sun Shin, Do-Guk Kim, Young-Duk Seo

TL;DR

The paper tackles the challenge of evaluating Korean writing quality in a timely and explainable way. It proposes UKTA, a Unified Korean Text Analyzer, with a three-stage pipeline: accurate morpheme analysis, mid-level lexical feature extraction across 294 metrics, and a rubric-based automatic writing evaluation using an attention-augmented model that fuses sentence-level KoBERT/BiGRU representations with essay-level features. On AI-HUB's Essay Evaluation Dataset, UKTA improves both accuracy and Quadratic Weighted Kappa ($QWK$) compared with a baseline, with improvements in 9 of 10 rubrics and informative feature-level explanations. The system outputs transparent rubric scores and top contributing features, enabling reliable and explainable feedback for Korean learners.

Abstract

Evaluating writing quality is complex and time-consuming often delaying feedback to learners. While automated writing evaluation tools are effective for English, Korean automated writing evaluation tools face challenges due to their inability to address multi-view analysis, error propagation, and evaluation explainability. To overcome these challenges, we introduce UKTA (Unified Korean Text Analyzer), a comprehensive Korea text analysis and writing evaluation system. UKTA provides accurate low-level morpheme analysis, key lexical features for mid-level explainability, and transparent high-level rubric-based writing scores. Our approach enhances accuracy and quadratic weighted kappa over existing baseline, positioning UKTA as a leading multi-perspective tool for Korean text analysis and writing evaluation.

UKTA: Unified Korean Text Analyzer

TL;DR

The paper tackles the challenge of evaluating Korean writing quality in a timely and explainable way. It proposes UKTA, a Unified Korean Text Analyzer, with a three-stage pipeline: accurate morpheme analysis, mid-level lexical feature extraction across 294 metrics, and a rubric-based automatic writing evaluation using an attention-augmented model that fuses sentence-level KoBERT/BiGRU representations with essay-level features. On AI-HUB's Essay Evaluation Dataset, UKTA improves both accuracy and Quadratic Weighted Kappa () compared with a baseline, with improvements in 9 of 10 rubrics and informative feature-level explanations. The system outputs transparent rubric scores and top contributing features, enabling reliable and explainable feedback for Korean learners.

Abstract

Evaluating writing quality is complex and time-consuming often delaying feedback to learners. While automated writing evaluation tools are effective for English, Korean automated writing evaluation tools face challenges due to their inability to address multi-view analysis, error propagation, and evaluation explainability. To overcome these challenges, we introduce UKTA (Unified Korean Text Analyzer), a comprehensive Korea text analysis and writing evaluation system. UKTA provides accurate low-level morpheme analysis, key lexical features for mid-level explainability, and transparent high-level rubric-based writing scores. Our approach enhances accuracy and quadratic weighted kappa over existing baseline, positioning UKTA as a leading multi-perspective tool for Korean text analysis and writing evaluation.

Paper Structure

This paper contains 12 sections, 1 equation, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Text analyzers
  4. Automated Writing Evaluation Tools
  5. Unified Korean Text Analyzer
  6. Morpheme Analysis
  7. Lexical Feature Analysis
  8. Automatic Writing Evaluation
  9. Experiments
  10. Experimental Settings
  11. Experimental Results
  12. Conclusion

Figures (2)

  • Figure 1: Illustrative overview of UKTA.
  • Figure 2: UKTA functionality. (A) Functionality in morpheme analysis results: Providing both table (A-1) and list (A-1) format, with an interactive and intuitive intuitive interface; results can be downloaded in JSON and TXT formats (A-3). (B) Functionality in lexical feature analysis results: Provided as categorized lexical features (B-1) with a list format (B-2); results can be downloaded in TXT and CSV format with selected features (B-3).