| 中文摘要 |
為了解決勞動力短缺與提高工作效率,近年來最新的科學技術,特別是AI(人工智慧)已被廣泛應用於各種領域。同樣地,台灣的大學教師也忙碌於教學、研究以及校務之中。本文旨在為了減輕教師的工作負擔,使用了Google公司開發並提供的對話型生成AI「Bard」(現稱「Gemini」),對於修改學生的日語作文能夠做到何種程度,可能會有何種問題進行了調查。為此準備了70篇短文(每篇35字以內)與2篇長文(約500-600字)進行測試。結果顯示,AI能夠以相對較高的準確性修正錯誤的用法,並且能夠將特定的單字在一定程度上轉換為適合文體的詞彙。然而,AI產生幻覺(hallucination)的情況並不少見,偶爾會給出虛假的結果。因此,使用者不可將AI輸出的結果照單全收,必須仔細檢查和確認。 In recent years, the utilization of cutting-edge scientific technologies, particularly Al (artificial intelligence), has emerged as a promising solution to address labor shortages and enhance work efficiency, finding applications across various domains. Similarly, university professors in Taiwan are often overwhelmed by their teaching responsibilities, research commitments, and administrative tasks. This study aims to investigate the potential of Google's interactive generative AI ''Bard'' in alleviating the workload of teachers. To begin, we evaluated Bard's capabilities and limitations in correcting Japanese compositions written by learners. The evaluation consisted of 70 short sentences and two long essays. The findings indicate that Bard can effectively correct errors with relatively high accuracy and, at the word level, can to some extent convert words to match the appropriate style of the text. However, it is crucial for users to carefully review and verify the output results, as AI can sometimes experience hallucinations and generate false information. |