ChatGPT是基于大规模数据训练的语言模型,其主要训练机制包括“下一个词预测” (Next-token prediction)、对齐 (alignment) 以及有人类反馈的强化学习机制 (Reinforcement Learning from Human Feedback, RLHF)。 本文从ChatGPT的工作原理出发,分析其在以当代中国文学与电影为选题的高年级中文课上的应用表现,包括课程设计、课堂讨论支持及对学生写作的反馈。本文认为, ChatGPT的工作原理决定了其在帮助学习者掌握词汇、语法、篇章方面的语言使用规则上可以发挥相当的作用。然而在中文学习的高级阶段, read more
This article uses Text World Theory (Werth 1999; Gavins 2007) in conjunction with VUE (Visual Understanding Environment) concept mapping software to analyze three statements from the trial of Amanda Knox, who was charged (along with her boyfriend, Raffaele Sollecito) with the murder of Meredith… read more
Addressing two hypotheses, the Aspect Hypothesis (Andersen & Shirai 1994; Shirai & Andersen 1995) and the Verb-Island Hypothesis (Tomasello 1992), this study analyzed whether the acquisition of third person singular present -s (3S) follows the same path as other tense-aspect markers, such as -ing,… read more