基于练习题目文本语义嵌入的知识追踪

打开文本图片集

关键词：深度学习；机器学习；教育领域的人工智能；自然语言处理；预训练语言模型；知识追踪中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2025）10-010-2972-08doi：10.19734/j.issn.1001-3695.2025.03.0086

Exercise semantic embedding for knowledge tracing

Cheng Zhia†，Li Jinlongb （a.SchooloftfeligeceelfeeceUestfeee， Hefei 230026，China）

Abstract：Toaddress theisseof por performanceofknowledgetracing modelsonnew exercises inopen-domain exercise scenarios，this studyinvestigated exercisesemanticembedding forknowledgetracing.The modelcomprised twostages：the firststageutilizedafine-tunedpre-trainedlanguage modeltoobtain exercisesemanticembedding，andthesecondstageemployed theembedingstocomplete knowledge tracingtaskswhile designingananswerencoder tocapture students’response information.To validate the model'sperformance，this paper conducted comparative experiments onanEnglish reading comprehensioneducationdataset，comparing themodel with state-of-the-artknowledge tracing models.Experimental resultsshow thatthe modelsignificantlyoutperforms comparative modelsinAUCandACC，effectivelyimproving the knowledgetracingperformance for new exercises inopen-domaindata.Theproposed methodachieves efective modelingof new exercises through exercise semantic embeddngand providesan innovativesolution fortheapplicationof intelligent education systems iopendomain data.

Keywords：deeplearning；machinelearing；artificial inteligence ineducation；natural language processing；pre-trained language model；knowledge tracing

0 引言

随着智能教育系统的蓬勃发展和大规模开放在线课程（MOOC）的广泛应用，教育领域正经历一场深刻的变革。（剩余21301字）