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

As an important learning method, online collaborative learning has been widely used by educators and learners in the field of education. Interactions are considered as the crucial feature of collaborative learning. There will be a large number of interactive texts during online collaborative learning. However, little study on how to analyze interactive texts automatically during online collaborative learning has been done so far. It is very difficult for detecting problems as well as providing real-time feedback and personalized learning support for learners due to deferred analysis, thus leading to low-efficient online collaborative learning. This study aims to investigate the interaction analytical key technology in online collaborative learning based on educational knowledge graph. First, this study will conduct interactive texts classification automatically by deep learning algorithm and interaction analysis method based on information flows during online collaborative learning. Second, entity extraction and entity linking will be conducted based on interactive texts during online collaborative learning in order to generate educational knowledge graphs of groups and individuals. Third, the degree of co-construction knowledge will be analyzed and evaluated automatically based on the educational knowledge graphs. The interactive text classification model and the method of entity extraction proposed by this study will provide insights into the latest progress and problems of online collaborative learning. The evaluation model will also shed light on whether or not learners achieve learning goals. The findings of this study will also provide important evidence on interventions and making decisions for teachers, thus improving online collaborative learning.

在线协作学习作为一种重要的学习方式被越来越多的教育者和学习者采用。交互是协作学习的基本特征，在线协作学习过程中将产生海量的交互文本。然而，鲜有研究对在线协作学习的交互文本进行实时分析。滞后的分析不仅难以发现学习者遇到的问题，而且无法为学习者提供实时反馈和个性化的支持，因此导致学习效果低下。本研究旨在探索基于教育知识图谱的在线协作学习交互分析关键技术，主要开展三方面研究：1.采用基于信息流的协作学习交互分析方法并结合深度学习算法对在线交互文本自动分类；2.对在线协作学习交互文本进行实体抽取和链接，从而生成小组和个体的教育知识图谱；3.基于教育知识图谱自动评价协同知识建构程度。本研究提出的在线协作学习交互文本分类模型、实体抽取和链接方法及评价模型不仅能够帮助教师及时掌握学习者的知识进展、而且能够评价学习者是否达到学习目标，也为教师干预和制定决策提供重要依据，从而切实提高在线协作学习的效果。

在线协作学习作为一种重要的学习方式被越来越多的教育者和学习者采用。交互是协作学习的重要特征和方式。在线协作学习过程中，学习者之间的在线交互产生了海量的交互文本。然而，鲜有研究对在线协作学习的交互文本进行实时分析。滞后的分析不仅难以发现学习者遇到的问题，而且无法为学习者提供实时的反馈和个性化的支持，因此导致学习效果低下。本研究旨在探索基于教育知识图谱的在线协作学习交互分析关键技术。三年来，本项目主要开展了如下五方面的研究工作：第一，深入探索并提出了数据驱动的协作学习的设计方法和优化技术；第二，采用基于信息流的协作学习交互分析方法并结合深度神经网络模型的算法对在线交互文本实现了自动分类，通过大量数据的检验表明模型的准确率高于其他基线模型；第三，提出并实现了基于学习分析技术对在线协作学习进行实时反馈的技术；第四，在不同领域实现了基于在线协作学习交互文本进行实体和关系抽取并自动构建小组和个体的教育知识图谱；第五，实现了基于教育知识图谱自动评价协同知识建构水平，通过开发的不同指标全面客观地评价协同知识建构水平。本研究提出的在线协作学习交互文本分类模型、领域知识图谱自动构建技术、基于教育知识图谱自动评价协同知识建构水平的技术不仅能够帮助教师及时掌握学习者的知识进展、而且能够评价学习者是否达到知识建构的预期目标，也为教师干预和制定决策提供了重要依据，从而切实提高在线协作学习的效果。