摘要
有效的在线协作学习可显著改善在线教学质量,而对在线协作学习过程的实时分析、监控和干预是促进协作学习行为有效发生的关键,这有赖于对在线协作学习交互文本的精准分类。为避免人工编码和传统机器学习方法分类效果欠佳的不足,采用基于深度神经网络的卷积神经网络(CNN)、长短时记忆(LSTM)、双向长短时记忆(Bi-LSTM)等模型构建面向在线协作学习交互文本的分类模型,以Word2Vec作为词向量,提出了包含数据收集整理、文本标签标注、数据预处理、词嵌入、数据采样、模型训练、模型调参和模型评价等步骤的在线协作学习交互文本自动分类方法。以知识语义类、调节类、情感类、问题类和无关信息类等作为交互文本的类别划分,对51组大学生所产生的16047条在线协作学习交互文本进行分类后发现:Bi-LSTM模型的分类效果最好,其整体准确率为77.42%;各文本分类模型在问题类、无关信息类交互文本上的准确率较低;CNN模型和LSTM模型在问题类交互文本上的分类效果更佳。该方法在面向在线协作学习的知识掌握度评估、学习活动维持、消极学习情绪干预、学习预警与提示等方面具有较高的应用价值。
The use of online collaborative learning can significantly improve the instructional quality in the field of E-learning.It is very crucial to conducting real-time analysis,monitoring,and intervention to promote the occurrence of online collaborative learning,which heavily relies on the precise classification of interactive texts.An interactive text classification model of online collaborative learning based on deep neural network is proposed in this study to overcome the insufficient classification of manual coding and traditional machine learning.This classification model adopted Word2 Vec as the word vector and it includes eight steps,namely collecting data,text annotation,data preprocessing,word embedding,data sampling,model training,model refining,and model evaluating.In the present study,51 groups of undergraduates are engaged in online collaborative learning and then16047 valid text data are obtained.These interactive texts are divided into five categories:knowledge semantic,regulation,emotional information,problems and irrelevant information.The three classifiers(CNN,LSTM,Bi-LSTM)are used for classification experiments.The results show that Bi-LSTM achieved the best classification results and the overall accuracy is 77.42%.However,the three classifiers achieved the low accuracy in terms of problems category and irrelevant information category.The CNN and LSTM achieved the higher accuracy than Bi-LSTM regarding the classification of problems texts.This method is very significant and valuable in terms of evaluating knowledge acquisition,maintaining learning activity,intervening in negative academic emotions,and learning alarming in online collaborative learning.
作者
甄园宜
郑兰琴
ZHEN Yuanyi;ZHENG Lanqin
出处
《现代远程教育研究》
CSSCI
北大核心
2020年第3期104-112,共9页
Modern Distance Education Research
基金
国家自然科学基金青年项目“基于教育知识图谱的在线协作学习交互分析关键技术研究”(61907003)
北京师范大学教育学部2019年度学科建设综合专项资金资助(2019QNJS010)。
关键词
在线协作学习
深度学习
深度神经网络
交互文本
文本分类
Online Collaborative Learning
Deep Learning
Deep Neural Network
Interactive Text
Text Classification
作者简介
甄园宜,硕士研究生,北京师范大学教育学部教育技术学院(北京100875);通讯作者:郑兰琴,博士,副教授,硕士生导师,北京师范大学教育学部教育技术学院(北京100875)。
