Earticle

초록

영어

This study examines linguistic changes in Japanese-language abstracts from KCI-indexed journals before and after the adoption of generative AI. A corpus of 582 abstracts by Korean researchers was divided into pre-AI (1990–2022, n=291) and post-AI (2025–January 2026, n=291) groups, with 2023–2024 excluded as a transitional period. Lexical analysis used log-odds ratios, and syntactic analysis employed Mann-Whitney U tests with rank-biserial correlation. At the lexical level, evaluative vocabulary foregrounding research contributions increased, while epistemic stance markers, the writer-reference term hissha, and critical engagement vocabulary declined substantially. At the syntactic level, genre-specific foundations remained stable, but structurally complex elements decreased selectively: clauses per sentence showed the largest effect size (r=−0.51), with concurrent reductions in sentence length, logical connectives, and ni yotte + reru/rareru constructions. The noun ratio rose significantly, yet genitive particle density declined, suggesting information compression through direct nominal compounding. These changes align consistently with known features of LLM-generated text across both lexical and syntactic levels, constituting evidence for the widespread involvement of generative AI in post-2025 L2 academic writing.

목차

1. はじめに
2. 先行研究と本研究の位置づけ
3. 分析の対象と方法
3.1 分析対象
3.2 分析方法
4. 分析結果
4.1 語彙選好の変化
4.2 統語構造の変化
5. まとめと今後の課題
【参考文献】

키워드

저자

• 李海渼 [ 이해미 | 慶熙大学 日本語学科 非常勤講師, 日本語学 ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  일본언어문화 [Journal of japanese Language and Culture]

• 간기

  계간

• pISSN

  1598-9585

• 수록기간

  2002~2026

• 등재여부

  KCI 등재

• 십진분류

  KDC 730 DDC 495