| 摘 要: 目前,传统的学业评价方法在反映学生的各项技能与知识掌握情况方面尚存一定不足,评价过程需要较多的时间与资源且难以实现个性化评价。文章首先探讨了ChatGPT(Chat Generative Pre-trained Transformer)在学业评价中的生成与应用,对学生学习数据进行诊断、激励、指导、干预。其次使用Bi-LSTM(Bi-directional Long Short-Term Memory)模型对评价文本进行情感分析,并使用BERT(Bidirectional Encoder Representation from Transformers)模型进行文本相似度检测,对ChatGPT评价内容与教师评价内容进行对比,结果显示:ChatGPT的评价内容情感在客观上更为积极,其评价内容文本相似度达到教师评价的75.21%以上,已具备实际应用价值与潜力。最后探讨了生成式AI在学业评价应用中的风险与启示。 |
| 关键词: 生成式人工智能;AIGC;学业评价;ChatGPT |
|
中图分类号: TP311.5
文献标识码: A
|
| 基金项目: 南京邮电大学教育科学“十三五”规划课题重点课题项目(GJS-XKT1901) |
|
| Research on the Application of Generative Artificial Intelligence in Academic Assessment—A Case Study of ChatGPT |
|
CHEN Siwen, KONG Yaqi, LIU Yu
|
(College of Education Science and Technology, Nanjing University of Posts and Telecommunications, Nanjing 210023, China)
1021163315@njupt.edu.cn; 1021163408@njupt.edu.cn; yliu@njupt.edu.cn
|
| Abstract: Currently, traditional methods of academic assessment often fall short in accurately reflecting students' mastery of various skills and knowledge. Its evaluation process requires considerable time and resources while struggling to achieve personalized evaluations. This research first explores the generation and application of ChatGPT (Chat Generative Pre-trained Transformer) in educational assessment, aiming to diagnose, motivate, guide, and intervene in students' learning progress using their academic data. Then, the study employs a Bi-LSTM (Bi-directional Long Short-Term Memory) model for sentiment analysis of evaluation text and utilizes a BERT (Bidirectional Encoder Representation from Transformers) model for text similarity detection. ChatGPT generated evaluations are compared with those provided by teachers. The results reveal that ChatGPT's evaluations exhibit more positive sentiment, with a text similarity of over 75.21% of the teacher's evaluation, demonstrating its practical applicability and potential. Lastly, the study examines the risks and insights associated with the implementation of generative AI in educational assessment. |
| Keywords: generative AI; AIGC(Artificial Intelligence Generated Content); academic evaluation; ChatGPT |
