学生学业述评智能生成模型的构建与实证研究  

The Construction and Empirical Study of the Intelligent Generation Model of Student's Academic Review

作　　者：熊余[1] 王盈[1] 蔡婷[1] 周松 蔡林沁[1] XIONG Yu;WANG Ying;CAI Ting;ZHOU Song;CAI Linqin(Chongqing University of Posts and Telecommunications,Chongqing 400065)

机构地区：[1]重庆邮电大学,重庆400065

出　　处：《现代远距离教育》2023年第1期32-39,共8页Modern Distance Education

基　　金：国家自然科学基金“虚拟学习环境类人情感交互与智适应研究”(编号:6227700);全国教育科学规划国家一般课题“人工智能与教育深度融合的政策体系研究”(编号:BGA210055);重庆市技术创新与应用发展专项重点项目“智能化教育评价关键技术研发与应用”(编号:cstc2021jscx-gksbX0059)。

摘　　要：教师每学期对学生进行学业述评是新时代教育评价改革的重要要求。但由于教师精力有限,常常难以对每个学生精准客观评价,导致撰写的学生学业述评存在模板化、公式化等问题。生成式人工智能技术能够有效利用学生的学业数据来实现述评的自动生成,为辅助教师开展个性化述评提供决策支撑。为此本研究充分挖掘学生学业个性化的认知特征,利用人工智能中的自然语言生成技术,构建面向学生学业述评的智能生成模型。首先设计偏科分析、进退步挖掘以及优差生判别等与学生学业认知相关的特征分析子模块,然后将特征分析结果融合原数据表内容,以输出准确且个性化的学生学业述评。通过在中国教育追踪调查系统数据集上的实证研究,证明设计的模型在各项指标上优于基线模型,达到了较好的性能效果。最后,为推进智能化的学业述评发展,建议要大力推动评价内容多维化、评价主体多元化以及评价技术多样化。It is a crucial requirement for the reform of educational evaluation in the new era that teachers conduct academic reviews of each student every semester.However,due to the limited energy of teachers,it*s often difficult to accurately and objectively portray each student,and there are problems such as templated and formulaic writing of students*academic reviews.Generative Artificial Intelligence can effectively realize the automatic generation of reviews using students'academic data,and provide decision support for assisting teachers to carry out personalized reviews.To this end,this research systematically combs the development status of natural language generation technology,and combines the cognitive intelligence characteristics of students*academic individualization to build a model for automatic generation of students,academic reviews.Firstly,the student data is processed through three feature analysis sub-modules,namely partial subject analysis,progress and backward mining,and discrimination analysis of outstanding and poor students.Then,the result of the feature analysis and the content of the original data table are jointly generated to output accurate and personalized student academic reviews.Through an empirical study on the China Education Tracking Survey System dataset,it is proved that the proposed model is superior to the baseline model in various indicators and achieves better performance.Finally,in order to promote the development of intelligent academic reviews,we should vigorously promote the multi-dimensional evaluation content,the diversification of evaluation subjects,and the variety of evaluation technology.

关 键 词：生成式人工智能 认知智能 学业述评 自然语言生成 个性化特征分析 

分 类 号：G43[文化科学—教育学]