研究生: |
黃昱涵 HUANG, YU-HAN |
---|---|
論文名稱: |
運用生成式 AI 協助學習者提升句子文法等級 GrammarUp: Applying Generative AI to Improve Sentence Grammar Level |
指導教授: |
張俊盛
CHANG, JYUN-SHENG |
口試委員: |
杜海倫
Tu, Hai-Lun 楊謦瑜 Yang, Ching-Yu 蕭若綺 Hsiao, Jo-Chi |
學位類別: |
碩士 Master |
系所名稱: |
電機資訊學院 - 資訊系統與應用研究所 Institute of Information Systems and Applications |
論文出版年: | 2025 |
畢業學年度: | 113 |
語文別: | 英文 |
論文頁數: | 47 |
中文關鍵詞: | 生成式人工智慧 、語言模型 、微調 、重述或釋意 、句法提升 |
外文關鍵詞: | Generative Artificial Intelligence, Language Model, Fine-tuning, Rephrasing or Paraphrasing, Syntactic Advancement |
相關次數: | 點閱:96 下載:6 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
本研究提出一套英文句子文法升級系統,能針對輸入句子生成文法結構更進階的改寫版本,且同時保留原句語意,並提供文法解釋。該系統結合生成式人工智慧技術與英語分級語法資源,專門為第二語言學習者設計,旨在促進其寫作能力的提升,亦可作為文法學習的輔助工具。我們依據英語分級語法資源建立訓練資料集,並對預訓練的語言模型進行微調,以實現語法升級的目標。初步評估結果顯示,此方法能有效生成具可行性且符合教學需求的升級句子,對提升學習者的寫作表現具有正向幫助。
We introduce a method for generating a syntactically more advanced sentence for a given sentence, while preserving its original meaning, and providing a brief grammar explanation. In our approach, pairs of basic and advanced sentences are generated, using a generative AI system and graded grammar resources for second language learning. The method involves constructing a training set of aligned sentence pairs across the Common European Framework of Reference for Languages (CEFR) levels, fine-tuning a pretrained large language model. At run-time, input sentences are passed through the fine-tuned model to generate syntactically more advanced sentences and corresponding grammar explanations to support learning. We present a prototype system, GrammarUp, that applies the method to a generative AI system and the English Grammar Profile (EGP). Evaluation of a set of learner sentences shows that the method generates reasonable results useful in a writing development setting. Our methodology cleanly supports combining generative AI and expert grammar knowledge, resulting in an innovative grammar tutoring system for second-language learners.