关键词:
人工智能
听力理解
试题开发
摘要:
本文对比了ChatGPT-4o和文心大模型-3.5生成初中英语听力文本和试题的能力,以2023年杭州市中考英语听力真题为样本,使用两种人工智能分别生成文本和试题。对文本和试题的分析结果表明,在生成听力文本上,ChatGPT和文心大模型差距较小,二者都能生成话题、词数、难度与样本文本基本一致的文本。在生成听力试题上,ChatGPT的表现好于文心大模型,相比前者,后者生成的试题出现了更多问题。笔者认为,鉴于寻找或编写听力文本的困难,英语教师在为学生编写低利害的练习题时,使用生成式人工智能来生成听力文本是一个不错的选择,但要生成高质量的试题,则需反复修改,以及需要更多人类智慧的参与。This paper compared the capability of ChatGPT-4o and Large Model ERNIE-3.5 to generate English listening comprehension texts and items for junior high school students, using the 2023 Hangzhou city Zhongkao listening texts and items as samples. Both artificial intelligence models were used to generate texts and items. The analysis results showed that in terms of generating listening texts, ChatGPT and Large Model ERNIE exhibited minimal differences, as both can produce texts with topics, word count, and difficulty similar to the sample texts. In generating listening items, ChatGPT performed better than Large Model ERNIE, with the latter producing more problematic items. The author argues that, given the difficulty of finding or creating listening texts, using generative AI to produce listening texts is a good option for English teachers when preparing low-stakes practice exercises for students. However, to generate high-quality test items, repeated revisions and more human engagement are necessary.